DevSwarm turns one prompt into a working full-stack app, and instruments the whole thing so you can see how it got there. Five specialist open-weight models take one role each: a planner locks a typed API contract, a frontend and a backend build against it in parallel, an independent critic reviews…

otel-swarm is OpenTelemetry instrumentation for multi-agent LLM systems, plus the SigNoz pack to read it. One createSwarm() call gives you a root span per run, a child span per agent, and a GenAI-semconv span per model call. Model fallbacks are recorded as span events carrying the model that…

Genvora is an AI-readiness auditing platform that shows businesses how AI assistants like ChatGPT, Gemini, Claude, and Perplexity actually see their website — not how a human sees it. It crawls a site simulating real AI crawler user agents (GPTBot, ClaudeBot, PerplexityBot, Google-Extended)…

ARGUS: AI Agent Runtime Governance & Observability Platform As autonomous AI agents execute increasingly complex, multi-step workflows, tracking their behavior, cost, and reliability becomes a black box. ARGUS is a comprehensive runtime governance and observability platform built to monitor…

An AI-powered Incident Triage Agent that simulates an on-call engineer's assistant during a production incident. Given an incident description, it plans an investigation, searches logs, checks a matching runbook, scores severity, and recommends an action — chaining two LLM calls (Google Gemini)…

Provenance is an AI agent that checks whether an invention is genuinely novel before you spend time or money filing a patent, and proves its answer instead of just asserting it. It searches real patent databases (Google Patents via BigQuery, and the EPO's Open Patent Services), synthesizes a…

Agent Flight Recorder extends SigNoz with an AI-powered investigation layer. It captures AI agent failures, replays executions, performs counterfactual analysis to validate fixes, identifies the root cause, and generates actionable engineering insights. Instead of just showing that an AI agent…

Whodunit turns root-cause analysis into a SigNoz query you own. You point it at a set of failing traces. It automatically picks a matched cohort of healthy traces, then mines the span, edge, and log patterns that actually separate "broken" from "fine." The winning pattern is compiled into a real…

Home APM turns your Home Assistant house into a distributed-tracing system on SigNoz. Home Assistant records a full trace of every automation it runs, but its built-in viewer shows them as unreadable node paths and keeps only about five runs, so the run you need is usually already gone. Home APM is…

Funnel of Thought measures an AI agent's reasoning contract — plan → tool → validate → respond — as a conversion funnel in SigNoz Trace Funnels. My agent has a validation step. A dashboard said it ran in 100% of runs. A funnel over the same 125 traces said 64%. Both queries were correct: a counter…

Yaatra is an AI-powered travel assistant that helps users plan personalized trips. The application uses a React frontend, a Node.js gateway, and a Python FastAPI AI engine to generate travel itineraries using AI. I integrated OpenTelemetry and SigNoz to implement end-to-end observability, enabling…

Dev-Rel-AI-Bot-Analytics This project automates GitHub issue triage using n8n, Gemini AI, and SigNoz. When a new GitHub issue is created, a webhook triggers an n8n workflow that sends the issue title and description to Gemini for classification. Gemini categorizes the issue (bug, feature request…

An offline-capable Python AI agent powered by Ollama, fully instrumented with OpenTelemetry to demonstrate local AI agent observability. The system features a root-and-child span hierarchy for request lifecycle tracking, tool invocation monitoring, dynamic attribute capture, and custom dashboarding…

AgentScope is recursive observability for multi-agent systems. A LangGraph pipeline (Planner → Researcher → Summarizer) is instrumented with OpenTelemetry GenAI semantic conventions into self-hosted SigNoz (Foundry + casting.yaml). A Watchdog meta-agent continuously polls the SigNoz MCP API to…

AgentNemesis is a trust-auditing dashboard for AI agents. It reads real OpenTelemetry traces from SigNoz and automatically catches four ways agents fail: getting stuck in unproductive loops, making unverified claims, breaking promises (saying they did something they didn't), and losing information…

SpanLedger is an out-of-band telemetry auditor for OpenTelemetry. It checks if traces, logs, and metrics actually reach SigNoz and helps detect silent telemetry loss before you trust your observability data.

ATC is an Air Traffic Control system for AI agents. It checks every tool call before execution, applies governance rules, and records every important decision for operators and engineers.

Vitals is an AI monitoring layer built on top of SigNoz. It detects behavior, quality, and cost changes in LLM applications and helps engineers find AI issues before users notice them.

Incident Observability Analyzer is an AI-powered workflow that automates production incident analysis using n8n and Google Gemini, while leveraging OpenTelemetry and SigNoz to provide end-to-end observability through traces and execution insights.

Sentinel Trace is a real-time security layer for AI agents built on SigNoz. Instead of only showing what an AI agent did, it helps determine whether its actions were safe and authorized. Most observability platforms can trace prompts, latency, and token usage, but they don't verify if an agent…

DevGuard AI is an autonomous, self-observing security agent that scans code for vulnerabilities (SQLi, command injection, and more), explains and fixes them using a multi-agent LLM pipeline (Router → Fixer → Validator with reflection loops), and — critically — watches its own health while doing it…

SpanSaver is an AI-observability agent: it traces your AI-native and infrastructure systems in SigNoz, detects where they leak money and tokens, proves a fix is safe, applies it, and verifies the result against live metrics — closing the loop. It even traces its own LLM calls into SigNoz, so it…

users overpay for expensive models when cheaper alternatives often produce equivalent output. mini_ghost turns Signoz from a passive cost tracker into an active cost optimizer.

I built Agent Drift Radar to catch something most observability tools miss - an AI agent that doesn't error out, but quietly starts behaving differently. More rambling answers, skipped verification steps, more retries, higher token costs, all while the content it produces still looks fine. My agent…

SRE Agent + Sidekick: a tool-calling AI support agent (Gemini + FastAPI) instrumented end-to-end with OpenTelemetry GenAI semantic conventions, shipping traces, metrics, and logs to SigNoz. A second agent, the SRE Sidekick, connects to SigNoz's own MCP server and uses that same telemetry to answer…

Preflight creates predictive execution contracts for AI coding agents before they start work. Instead of only reporting cost or scope drift after it happens, it predicts the appropriate model tier, effort, and expected file footprint up front, then compares the agent’s observed behavior against…

The SD-WAN Enterprise NOC is an AI-powered network monitoring and observability platform that simulates a real-world Enterprise Network Operations Center (NOC). It continuously monitors WAN infrastructure, analyzes network health, predicts failures, automatically generates incident tickets, and…

Project Entropy deliberately injects controlled failures into your AI agent's environment then automatically checks how well your agent recovers. It turns unpredictable production crashes into measurable data you can act on.

GPUs are expensive and unevenly distributed. A developer might have a Mac with no GPU, while a friend or teammate has a Windows laptop with an NVIDIA GTX 1650 sitting idle. Running AI workloads, model inference, or even simple tensor computations becomes impossible without local GPU access.

Sentinel doesn't just tell you something's broken, it fixes it, proves the fix worked, and writes the postmortem before your coffee gets cold. Point it at a live SigNoz alert and it investigates root cause, proposes a fix, waits for your explicit yes, applies it, verifies recovery against real…

CodeAutopsy is git blame for the AI era. As developers increasingly rely on AI agents to write code, production crashes leave behind an attribution gap: standard telemetry shows where code failed, but not why the AI wrote it that way. CodeAutopsy bridges the build/run boundary by capturing the AI’s…

Ouroboros is an AI SRE agent whose only senses and hands are the SigNoz MCP server. It detects anomalies via SigNoz alerts, root-causes them by querying traces, logs, and metrics through MCP, executes a remediation, and verifies the fix, all while being fully instrumented into SigNoz itself, so you…

Sentinel — a self-healing trading bot with an AI SRE agent, fully observed in SigNoz. Sentinel is a paper-trading bot (real Binance prices, simulated money) that observes itself end-to-end through SigNoz and heals itself when its data source fails. When Binance starts erroring, a local circuit…

AI Agentic prototype this helps our district issue's management with quick response and secure all problems at a place where Magistrate can review work and track the progress.

Signoz Hermes SRE Agent: Signoz Alert → Hermes → Telegram SigNoz alert wakes Hermes, Hermes does RCA on live telemetry, then messages you on Telegram so you can verify, steer, and approve a fix.

Tower Trace is a policy + budget control kernel for tool-using AI agents. It intercepts every tool call, then allows, denies, or hard-kills on budget breach. Built for Agents of SigNoz Track 01. Demo shows happy path, injection deny, and runaway budget kill, with full proof in SigNoz.

AI agents are increasingly allowed to act on their own — calling tools, updating records, sending messages — without anyone approving each individual action. The risk this creates is easy to overlook: if an agent gets manipulated into doing something outside its intended scope, the response it…

mcp-observer is an OpenTelemetry-native transparent proxy and AI Diagnostic Assistant for Model Context Protocol (MCP) servers, powered by SigNoz. THE PROBLEM: MCP has no native observability. When an AI agent retries a broken tool 50 times, hangs on a deadlocked server, or encounters dynamic tool…

Agent K is a closed observability loop. A small instrumented AI app (Orbit) emits OpenTelemetry traces, metrics, and logs to SigNoz. When a golden signal crosses a threshold (error rate, p95 latency, LLM cost, or retrieval relevance), a SigNoz alert webhooks Agent K — an LLM agent that investigates…

An AI agent can finish every step and still return the wrong answer. Agent Flight Recorder, or AFR, turns that kind of failure into a tested draft pull request while keeping publication under human control. The demo starts with a real multi-agent bug. A researcher finds a fact and its source, an…

Argus is a persistent web monitoring and research agent designed to track topics that change over time. Users create a “pursuit” by defining a topic, objective, schedule, source rules, and the record fields they want to maintain. Argus then searches public sources, retrieves relevant content…

# Nights Watch: Guarding AI Agents Beyond the Wall AI agents rarely fail by crashing. They fail by quietly succeeding at the wrong task. A travel agent asked to book a flight under $400 may confidently purchase a $1,200 ticket and report "task completed." Nights Watch is a runtime resilience layer…

AI Incident Investigator is an AI-assisted incident investigation platform that turns observability data into actionable root-cause reports. Our FastAPI application is instrumented with OpenTelemetry to generate traces, metrics, and trace-correlated logs, which are collected and visualized in…

FleetOps is a fleet-observability pipeline built on top of Euro Truck Simulator 2. A Python daemon polls the game's telemetry API and pushes metrics, logs, and traces to a self-hosted SigNoz instance, treating each delivery job as a distributed trace so a dispatcher can see exactly where and when…

OSS First Mate is a live AI assistant that helps open source maintainers manage their GitHub issue backlog. It fetches issues through Coral's SQL interface, classifies them with Groq's LLM, finds duplicates, and drafts release notes. For this hackathon, I instrumented the whole pipeline with…

Gradebook makes "our AI is doing well" a measurement instead of an assertion. Every AI decision it wraps emits a standard OpenTelemetry gen_ai.evaluation.result event, and every grade on that event carries one extra field, augmentloop.grade.source, saying where its authority came from. math means a…

TemporalGuard is an agent-native business workflow observability platform powered by OpenTelemetry and SigNoz. It helps teams monitor processes that may begin successfully but fail to reach their expected outcome, such as a document being uploaded but never verified, an employee being offboarded…

Teams now run several coding agents at once, Claude Code, Codex, OpenCode, Gemini CLI — and whoever owns that rollout can't answer basic questions: what did agents cost this week, per developer, per repo? How many sessions looped for twenty minutes and produced nothing? Can we take a degraded MCP…

AgentWatch is a reusable, open-source observability layer for autonomous AI agents. Wrap your agent with @observe and every LLM call and tool call becomes an OpenTelemetry trace - with token cost, latency, and success rate - that lights up SigNoz natively via the GenAI semantic conventions. It even…

Axiom AI is an autonomous backend performance engineer that helps developers find, validate, and implement performance optimizations with minimal manual effort. Developers connect a GitHub repository, and Axiom automatically discovers API endpoints, instruments the application with OpenTelemetry…

AgentTrail is a real-time security and data-lineage layer for AI agents, built on SigNoz. While using Claude for everyday office work, I noticed a file had changed that we never touch — if I hadn't caught it right then, it could have been shipped. There was no good way to see exactly what commands…

Mirage is an AI Forensic Investigation Platform that extends traditional observability to AI systems. Built for the Agents of SigNoz Hackathon, it combines OpenTelemetry, SigNoz, the Model Context Protocol (MCP), and a custom Trust Engine to investigate every AI interaction. A live Northwind…

Yhh

Sentinel is an AI Runtime Safety Supervisor built on SigNoz and OpenTelemetry. It transforms traces, metrics, and logs into evidence-backed runtime intelligence for autonomous AI workflows. Instead of only showing what happened, Sentinel evaluates whether an AI workflow is operating safely…

VibeCheck is a self-learning prompt optimization layer for AI coding assistants. The Problem: Vibe coding — building software by prompting an AI instead of writing code by hand — has become the default workflow for many developers. But prompt quality directly determines token usage and cost, and…

Rewind is a flight recorder and replay system for AI agents — regression testing where real production incidents become the test suite. It records every agent step (LLM calls, tool calls, full payloads) as OpenTelemetry GenAI spans into self-hosted SigNoz. Take any trace id and rewind replay…

Aegis-ZeroEgress-RCA is a self-hosted, local-first incident diagnosis system that uses a small open-source model and a restricted observability tool interface to analyze telemetry and explain root causes without sending sensitive operational data outside the environment.

Track: 1 — AI & Agent Observability. That's WarrantTrace's actual scope — detecting a violation in how an agent's authority gets exercised, not just visualizing signals or a custom side-project (Track 3) or an OTel instrumentation exercise (Track 2). It does touch Track 2 heavily (Query…

AegisMesh is an Autonomous Multi-Signal AI SRE Controller that closes the operational observability loop for cloud-native infrastructure outages. Built for Track 01: Agents of SigNoz, AegisMesh transforms passive SigNoz telemetry into an active, safe, self-healing control plane. 🛑 The Problem In…

We built an insurance claims bot, then asked an uncomfortable question: if it starts confidently making things up, would our monitoring even notice? It wouldn't — latency, errors, and cost all stay green while the answer is a lie. Homeostat fixes that. It grades every answer against the documents…

Non-Determinism Regression Catcher is an observability-driven framework for detecting silent regressions in LLM applications. Instead of evaluating a single model response, it repeatedly executes the same task, instruments every execution with OpenTelemetry, and exports traces and custom metrics to…

AI agents fail silently. Our demo support agent, asked to fix a duplicate ₹2,499 charge, refunded ₹4,998 - every HTTP call returned 200, nothing crashed, and classic monitoring called the system healthy while it lost money. Trace2Test is a control plane that makes such failures happen exactly once…

Sentinel AI is an autonomous SRE and self-healing infrastructure platform for Kubernetes applications. It continuously ingests live telemetry from a multi-service application, correlates service health, latency, error rate, logs, traces, and Kubernetes deployment state, then creates an incident…

AXRAY is a flight recorder for autonomous AI coding agents, built on OpenTelemetry and powered by SigNoz. Autonomous coding agents don't behave like traditional services , a single turn might involve an LLM generating a plan, followed by multiple tool calls inside a sandbox, each with its own…

Jargons is an AI code-review agent for GitHub — and, unlike most review bots, it's fully observable. Install the GitHub App, and Jargons automatically reviews every pull request: it reads the diff, flags bugs and security issues with severity and file locations, and can open a companion pull…

An autonomous, closed-loop SRE and telemetry engine that monitors runtime traces, detects schema drift and performance anomalies via SigNoz, and dynamically patches system failure modes in real time.

Cognivern is an autonomous trading agent that runs sealed-bid auctions on the Canton Devnet for tokenized oil futures. Every cycle, the agent calls an LLM, picks a bid, signs it, and submits it to a Daml contract — all without a human in the loop. The hard part was not the bidding: it was knowing…

Our project was an AI powered interview platform built to make the hiring process faster, smarter, and more consistent. A recruiter can upload a list of candidates, and the system automatically sends each applicant a personalized interview link by email. Before the interview begins, the platform…

AIIMS - All India Institute of Machine Sanity: a psychiatric hospital for AI agents. AI agents rarely crash cleanly - they go insane. They call the same tool with identical arguments eighteen times, retry-storm a dead dependency, or fill their context until they've forgotten the task and…

Contextline is a SigNoz-native incident-response agent for Kubernetes. It gathers live infrastructure evidence, diagnoses failures with an LLM, validates proposed remediations through deterministic safety checks, requires human approval for risky actions, and verifies whether the system actually…

Astrid uses SigNoz to collect metrics about your network traffic — every process, every destination, every byte. When something unusual happens, SigNoz sends an alert to Astrid's AI. The AI explains what's happening in plain English and suggests a fix. You can click a button to apply the fix…

TracePilot AI is an AI agent observability platform that traces every stage of an AI research agent's execution using OpenTelemetry and SigNoz. It computes a real-time health score, auto-detects performance bottlenecks, analyzes failures with root-cause explanations, tracks token usage and…

InterviewOps is an enterprise-grade technical interview preparation platform built specifically for Systems Engineers, Site Reliability Engineers (SREs), and Software Architects. By evaluating a candidate's resume against specific Job Descriptions (JDs), InterviewOps deploys a multi-agent AI…

Kernel-level truth for AI agents: it watches what the agent did, not what it reported,See what your AI agents really do — not what they claim — with zero instrumentation.Zero-instrumentation observability and enforcement for AI agents, captured at the kernel where agents can't hide.

Sentinel is a self-healing AI observability system built for Track 01. It has two agents - a Worker agent that handles user questions and is fully instrumented with OpenTelemetry traces, metrics, and logs sent to SigNoz, and a Sentinel agent that monitors that telemetry via the SigNoz MCP server…

AgentOps Sentinel is an AI-powered observability platform for LLM and multi-agent applications. It uses OpenTelemetry and SigNoz to collect traces, monitor performance, and detect failures across AI workflows. When an issue occurs, the system analyzes the telemetry, identifies the likely root…

Dynaland is a local-first observability platform for AI coding agents that brings production-grade monitoring to local development workflows.

Sentinel AI is an AI-powered incident analysis platform that integrates OpenTelemetry, SigNoz, PostgreSQL, and AWS Bedrock to automatically correlate logs, metrics, and traces, generate incident timelines, and provide AI-assisted root cause analysis for faster production troubleshooting.

GhostTrace AI is an AI-powered investigation and observability platform built to monitor, trace, and analyze LLM agent workflows in real time. It simulates real-world AI incidents such as slow APIs, hallucinations, token spikes, and tool failures while collecting traces, logs, and metrics using…

DebugProof is an incident-centered release gate that proves whether a new software release will still be understandable when it fails. It injects the same controlled failure into a known-good baseline and a candidate release, sends their OpenTelemetry traces, logs, and metrics to SigNoz, and checks…

Every AI agent demo shows what the agent did. We built something that checks whether you can trust what it told you. The scariest agent failures aren't crashes — they're silent: a dropped token count, a truncated context window, a broken span tree, with no error anywhere. Telemetry Health Guardian…

My project is all about the way to understand/ to watch something break in real time. National exam-results days are the perfect real-world scenario for this: massive predictable traffic spikes, panicked users, and a portal that has genuinely gone down in front of millions of people more than once…

CortexSOC is an autonomous AI-powered Security Operations Center (SOC) where 8 specialized agents (Log Collector, Threat Detection, MITRE Mapper, Investigation, Risk Scorer, Patch Recommendation, SOAR Executor, and Incident Report) collaborate to detect, investigate, and remediate cyber threats in…

Meridian follows a single payment from a fiat rail to real on-chain settlement — as one distributed trace, in SigNoz. When a fiat-to-crypto payment stalls, support usually can't tell which of five or six separate systems it died in. Meridian instruments the whole pipeline (payment processor…

AI-powered photo management platform that connects to Google Photos and local storage directories. The solution should identify duplicate and near-duplicate images, automatically categorize photos, detect important documents, and group people based on facial recognition.

An autonomous, self-healing LLM observability loop that uses OpenTelemetry, SigNoz, and local SLMs to detect attacks and hot-patch agent runtime security in real time.

Bachat+ is an AI-powered fintech platform that turns everyday spending into effortless investing. It automatically rounds up digital payments to the nearest ₹10 and invests the spare change into a Smart Investment Wallet, helping users build wealth without changing their spending habits. The…

AgentForge turns a one-line product idea into a complete software blueprint — requirements, architecture, database schema, backend API surface, frontend structure, and testing strategy — generated live by a 6-agent AI pipeline (powered by Groq's llama-3.3-70b-versatile) and streamed to the user in…

AgentLens makes AI agent quality a first-class observability signal. It instruments an AI customer-support agent with OpenTelemetry, ships the traces to SigNoz, then reads those finished traces back out of SigNoz to grade what the agent actually did writing the scores back in as OTel metrics. A…

NETRA AI We got tired of AI fraud models that give you a verdict and nothing else. Flagged or not flagged, and that's the end of the conversation — no idea what happened in between, no way to tell if the model's degrading or the pipeline's just slow. So NETRA AI is built around the opposite idea…

SRE Sidekick is an AI SRE agent that sits on top of a stock SigNoz deployment and runs the whole incident loop: detect, ground, diagnose, communicate, act, verify. Its headline feature is not that it explains outages. It is that it refuses to explain them when the telemetry cannot be trusted. The…

Agent K is a code-enforced incident-response agent for a FastAPI RAG support service. When a SigNoz alert fires, it investigates through the SigNoz MCP server, publishes a root-cause report, and can roll the service back — but only when plain Python says it may. It is built around three laws, each…

Overview Triage is a Security Operations Center built around OpenTelemetry for the SigNoz Hackathon. While most teams use OpenTelemetry to watch server health and latency, this project uses it to watch hackers. It stops treating security threats as generic server errors and instead tracks exactly…

Second Signature is an observability system for high-risk actions performed by AI agents. The demo reconstructs a CAD 12,500 refund to a new beneficiary under a policy requiring human approval for transfers above CAD 5,000. It compares three outcomes: a legacy agent that executes the payment…

Kams is an open-source observability and control boundary for the Model Context Protocol (MCP). It sits between an AI agent and its MCP servers, pins the exact tool definitions that were approved, and continuously checks for integrity drift, prompt-injection patterns, sensitive data exposure, cost…

FlightRules turns SigNoz traces into deterministic release contracts for AI agents. An agent can return the correct final answer while taking an unsafe execution path. In our demo, two refund-agent versions both tell the customer, “Your refund has been processed.” The approved version retrieves…

A self-hosted SigNoz observability stack deployed via Foundry (casting.yaml/casting.yaml.lock), instrumented end-to-end with a custom Node.js OpenTelemetry SDK. 𝐃𝐞𝐩𝐥𝐨𝐲𝐦𝐞𝐧𝐭 : SigNoz installed via Foundry (foundryctl) as a Docker Compose stack on an AWS EC2 instance — ClickHouse (telemetry store)…

Nexora is an all in house project where it manages sales, production planning, procurement, inventory, payroll, banking and 17 integrated modules for different business tracks all in sync for the VERP (Village Enterprise Resource Planning) System with Authentication, Authorization, and Role-Based…

TraceMind AI is an AI observability copilot built with FastAPI, React, Gemini AI, OpenTelemetry and SigNoz. It analyzes traces and telemetry, detects root causes, visualizes performance and provides AI-powered recommendations for faster debugging.

SigNoz Governor: An agent observability system that monitors your AI Agents in real time, and stops them when they trip or go wrong. Whenever your AI Agent is stuck doing something in a loop, or if its exceeding the budget set for AI Usage, the governor will pause the session right away preventing…

Cerberus is an AI SRE copilot for AI agents. It instruments an agent workload into SigNoz over OpenTelemetry, then reads that telemetry back through the SigNoz MCP server, ranks every run by failure and spend, and a copilot explains the worst incidents in plain English; grounded only in…

TraceMind AI is an AI-powered Site Reliability Engineering (SRE) assistant built with SigNoz and OpenTelemetry. The project helps developers monitor application health, analyze performance issues, and understand incidents using real-time observability data. It provides a dashboard with system…

AtlasAI is a multi-agent RAG assistant for industrial teams — it answers equipment questions using a plant's own manuals, maintenance logs, and incident history instead of generic search. A FastAPI backend routes queries through five specialized agents: a knowledge agent that grounds LLM answers…

Glasswing is a runtime that authorizes AI-agent tool calls, and SigNoz is its observability backbone — everything is emitted OpenTelemetry-native over OTLP and read in SigNoz across traces, logs, metrics, alerts, and dashboards:

Modalnos is a runtime audit layer for AI-generated code. It captures Claude Code edits and correlates changeId → runId → traceId → incident. SigNoz detects an N+1 regression from 21 real PostgreSQL spans. After optimization reduces it to 2 spans, the incident stays open until SigNoz confirms the…

Goose detects silent AI quality failures — cases where a RAG returns HTTP 200 but the answer is wrong. It instruments the app with OpenTelemetry quality metrics in SigNoz so retrieval misroutes and ungrounded answers show up as cliffs in quality.score, not just latency or error rate. When quality…

An AI-powered subtitle generator built with Next.js that processes video content seamlessly. To ensure complete reliability, performance tracking, and live monitoring, the application is integrated with OpenTelemetry and a self-hosted SigNoz observability stack managed locally via Foundry…

Over-Watch is an autonomous SRE agent that investigates production incidents on SigNoz the way a senior on-call engineer would — forming hypotheses, following the evidence through real traces and logs, and never guessing. It's built as a direct tool-calling loop (no LangGraph, no fixed DAG)…

Built Incident Autopilot, an AI-powered, approval-gated Kubernetes autoscaler that uses SigNoz observability data to detect incidents, analyze root causes, recommend scaling actions, and execute them only after human approval, ensuring safe and reliable production operations.

Sniper is a genuine, working agentic network-penetration-testing platform: a real LangGraph StateGraph drives an eight-stage pipeline, API backed-backed LLM narrates findings under strict anti-hallucination guardrails. 24 Dockerized open-source tools (nmap, nuclei, ZAP, Metasploit, netexec, and…

MIB (Men in Backend) is an autonomous AI-powered Site Reliability Engineering platform built on top of the SigNoz Model Context Protocol (MCP). Instead of acting as a chatbot for observability, MIB deploys two specialized AI agents with clearly defined responsibilities, deterministic workflows…

Shadow AI is one of 2026's fastest-growing security risks — nearly every organisation has employees running AI tools with no oversight, yet almost none have visibility into where that AI actually lives. Shadow AI Radar closes that gap by pointing observability at the machine itself. It's a Python…

Nervous System is a self-healing observability platform for AI agents. It uses SigNoz and OpenTelemetry to monitor agent health in real time, diagnose failures with an LLM, and automatically recover through actions like restarting, throttling, or rerouting unhealthy agents.

InterviewIQ AI is an AI-powered technical interview coach that helps candidates practice interviews and receive intelligent feedback on their answers. The platform generates an interactive interview experience where users answer technical questions and receive AI-based evaluation including scores…

Evolvex is an AI-powered Investigation OS built on top of SigNoz. SigNoz detects incidents (traces, logs, metrics, alerts) — Evolvex answers "why" by correlating that alert with everything that changed around it: Git commits, deployments, Kubernetes events, feature flags, and CI/CD events. When a…

A personal observability dashboard on top of self-hosted SigNoz: SigNoz is the metrics store + dashboard UI, and a small local collector pushes Screen Time and active-project data into it via OpenTelemetry (OTLP).

Ask your app what happened to it: an agent investigates your live SigNoz telemetry and returns the root cause, the evidence traces, and a fix from a plain-English question or straight from a firing alert.

AgentScope Sidekick is an AI-agent observability and incident investigation product for Track 1. It follows each agent request across retrieval, tool, LLM, API, and persistence spans; correlates traces, metrics, logs, and alert breaches; identifies a deterministic evidence-backed root cause…

TraceForge is an autonomous pre-production reliability agent that tests whether a code change is actually safe before deployment. It analyzes a Git diff, identifies affected API endpoints, generates a deterministic k6 load test, and runs baseline and candidate experiments. TraceForge then queries…

"Stuck" is a VS Code extension that provides observability for AI coding agents to solve the "black box" problem. It traces file edits, terminal commands, git commits, and internal agent tool calls as OpenTelemetry spans, and sends them to a local SigNoz instance. When a coding session finishes or…

Castellan is the trust layer for autonomous AI agents. It instruments AI agent workflows with OpenTelemetry, streams telemetry into SigNoz, and governs agent behavior in real time. Three engines work together: - Sentinel detects behavioral anomalies (tool loops, cost spikes, schema hallucinations…

AgentOps is an autonomous SRE Sidekick and Control Tower designed for production multi-agent AI systems. Built natively on SigNoz, OpenTelemetry GenAI Semantic Conventions, and the SigNoz MCP Server, AgentOps monitors complex agent workloads to solve unpredictable failure modes—including payment…

An AI agent that connects to a self-hosted SigNoz instance through its MCP server, answering natural-language questions about live observability data — service lists, traces, telemetry — grounded entirely in real data, with no invented responses. What sets it apart from a plain Q&A agent: every…

GamePulse AI is a futuristic AI-powered gaming intelligence platform that combines an immersive next-generation gaming interface with real-time AI agent monitoring and observability. The project introduces an intelligent gaming companion that analyzes player behavior, detects unusual events…

ObserveAI is a SigNoz-powered demo that turns telemetry into an AI-driven incident investigation assistant. It uses traces, metrics, logs, and service topology to automatically identify root cause, explain what happened, and recommend a fix all inside a transparent SigNoz workflow.

PulseGuard : A Java-based production incident observability platform that uses OpenTelemetry and SigNoz to detect, visualize and investigate application failures across metrics, logs, traces, dashboards and alerts.

AI SRE Copilot is a Streamlit app that analyzes application logs using Groq's Llama 3.3 70B to identify root cause, severity, and recommended fixes, instrumented end-to-end with OpenTelemetry and SigNoz for full observability into the AI layer itself.

GreenLight turns failing GitHub Actions into something that can almost fix itself. CI failures show up as OpenTelemetry traces in SigNoz. An alert fires to our agent, which digs into those traces via SigNoz MCP, patches the repo with GitHub MCP, and opens an autofix PR. A green Actions re-run is…

DriftWatch is a self-observing AI agent SDK. It instruments every tool call and model step an AI agent makes as OpenTelemetry, then runs an LLM over that telemetry to detect behavioral drift. It detects shifts in tool-call mix, error rate, latency, or token spend that don't throw an exception but…

Execution Behavior Intelligence (EBI) is an observability intelligence platform that helps engineers detect and understand runtime behavior changes across software deployments. Built with FastAPI, React, PostgreSQL, and Docker, the platform leverages OpenTelemetry and SigNoz to collect and analyze…

A durable, memory-enabled AI on-call SRE sidekick built on SigNoz - Agent99

SigNoz SRE Copilot is an autonomous self-healing infrastructure system built on SigNoz MCP Server and CrewAI. When a SigNoz alert fires, a multi-agent CrewAI system autonomously diagnoses the incident using SigNoz traces, metrics, and logs via the MCP Server, executes safe Kubernetes remediation…

Workday Integration Doctor is an AI agent that reads SigNoz telemetry via the MCP server and diagnoses enterprise integration failures like a senior consultant. A mock Workday tenant runs a multi-step provisioning orchestration instrumented with OpenTelemetry; config-driven chaos injection…

According to research from Microsoft, roughly 1 in 5 incidents in real production systems closely matches something that’s already happened before, and most teams have no systematic way to surface that match. The fix ends up living in someone’s memory, a Slack thread, or a stale runbook, while the…

Trakr is a Discord bot that monitors other bots in a server and sends their telemetry to Signoz using Opentelemetry. It tracks bot uptime, message activity, downtime, and moderation actions, helping server owners understand whether their bots are actually working instead of simply appearing online.

This project is an AI-driven observability platform designed to monitor, analyze, and improve the performance of AI agents and autonomous systems in real time. The platform collects logs, API traces, prompts, responses, execution metrics, and system events from AI agents, then uses machine learning…

Agent K asks a simple question: what happens when your AI agent is failing but every dashboard says it's fine? That's the actual failure mode with LLM agents. HTTP 200, normal latency, nothing paging anyone, while the agent is quietly hallucinating, citing the wrong chunk, or picking the wrong…

SignozForge Observer is a self-healing multi-agent AI system where an SRE Copilot monitors a 4-agent pipeline (Coordinator, Researcher, Coder, Reviewer), detects bad output using 7 independent anomaly detectors, and repairs failures automatically with a 2-strike healing path. SigNoz serves as the…

Overview Triage is a Security Operations Center built around OpenTelemetry for the SigNoz Hackathon. While most teams use OpenTelemetry to watch server health and latency, this project uses it to watch hackers. It stops treating security threats as generic server errors and instead tracks exactly…

Noval multi Agent Orchestration - Cursor, but for video editing. Drop in your raw clips, describe the video you want in a sentence, and get back a finished cut — captioned, scored, and colour-graded. We built an entire video editor from scratch either AI chat feature - which can edit, add…

CasperAgent is a live, multi-tenant AI meeting assistant (schedules bots, records calls, writes minutes, reads team dynamics) that is instrumented end-to-end into SigNoz — and then closes the loop by observing itself. It is not a toy instrumented for a demo. Every agent run, LLM call, tool…

Ouroboros AI is a self-healing observability system for AI agents. A simulated agent runs continuously and deliberately fails with context overloads; every step is traced with OpenTelemetry straight into SigNoz. SigNoz detects the failure pattern and fires a webhook automatically, which triggers an…

Autonomous SOC Observability Engine is a next-generation security operations platform built to eliminate alert fatigue and manual log triage. Powered by OpenTelemetry for high-throughput telemetry ingestion and a robust SigNoz and ClickHouse backend, the platform processes real-time security…

An AI agent that scrapes job listings from RemoteOK, summarizes them using a local LLM (Ollama/llama3.2), and stores results — fully instrumented with OpenTelemetry and observed via SigNoz. Includes custom trace attributes for token usage and model info, a 4-panel dashboard (duration, tokens, jobs…

FireDrill is an AI red team for your infrastructure. Chaos engineering asks whether your system survives a failure; FireDrill asks whether anyone would have found out. Five agents run a loop: recon learns your service graph from traces, a planner designs experiments and predicts which alert should…

AgentATC is a three-agent pipeline — Planner, Executor, and Critic — that collaborates to write a report, built for the "Agents of SigNoz" hackathon to expose a failure class standard observability can't see: coordination breakdowns between agents. A task can bounce back and forth forever, stall…

AI Council allows multiple specialized AI agents to deliberate, debate, and vote on complex tasks before presenting an approved action plan to the human. Once confirmed, an autonomous action engine executes the task. (this works for both abstract as well as practical work situations for the…

Missio is an evidence-bound remediation mission control system for production incidents. A real incident enters through Slack. Missio queries SigNoz for the corresponding OpenTelemetry trace, identifies the exact code location and deployed Git revision, and gives only that evidence-linked source to…

AlphaTrace bridges the gap between complex quantitative finance pipelines and modern operational visibility. It transforms a 7-stage ML/DL stock prediction system into a fully transparent, highly observable product. Instead of treating AI trading algorithms as black boxes, AlphaTrace gives quant…

AgentOps is a four-agent swarm that autonomously detects, diagnoses, fixes and verifies a real bug in a third-party open-source app — with no human in the loop. Monitor discovers the incident by querying SigNoz through the SigNoz MCP server, Diagnosis identifies the exact failing line from traces…

AegisSRE is an enterpise-grade, governed autonomous systems that combines observability, AI reasoning, and governed automation to resolve production incidents with minimal human intervention.

Agent Debugger is an AI agent built with Python and Groq (LLaMA 3.1) that answers questions in 3 steps: plan, search, and synthesize. Every step is fully instrumented with OpenTelemetry and sends traces to SigNoz. This makes the agent completely observable — you can see exactly how long each step…

TraceCheck is an open-source testing platform for AI agents where OpenTelemetry traces stored in self-hosted SigNoz are the ground truth. Instead of grading only an agent's final reply, TraceCheck runs white-box assertions against the actual trace SigNoz recorded — which tools were called, in what…

Notch is mission control for a fleet of coding agents. Claude Code, Codex, OpenCode, Grok and the Antigravity CLI each keep their own memory in their own files and run blind to each other. Notch gives them one shared brain and one baton: memory, decisions and context become a single thread that…

Every codebase has a DNA. CoDNA is an AI-powered repository intelligence platform that helps developers understand large software projects beyond individual files. Instead of treating a repository as a collection of disconnected source code, CoDNA builds a structured understanding of its…

Argus is a drop-in, OpenAI-compatible gateway built to catch AI-agent failures that normal infrastructure monitoring misses. Every request can return 200 OK, latency can look normal, and the agent can still be giving users the wrong answer. Argus handles this with two loops: PREVENT and LEARN…

The world is changing from clicking to typing, every company on this planet will employ an AI Chat bot, an AI Voice bot in their services but how do they kneo what failed what the agent uncovered and how do you convert sematic relationships to Insight, failures and debug the agents on what failed…

Zooid is an open-source instrumentation package that brings real OpenTelemetry observability to real-time voice AI agents, catching exactly where a delay happens, STT, the model, or TTS, something standard HTTP tracing can't see. pip install zooid gets the full engine, one line…

TraceBill bills your customers for the API calls they make. You add three lines of code to your app and it starts counting, pricing and invoicing on its own. The difference from other billing tools: you can click any charge and see the exact requests behind it.

Kams is a SigNoz-native observability and containment layer for the Model Context Protocol (MCP). It sits transparently between AI agents and MCP servers, requiring no agent-code changes. Kams emits OpenTelemetry traces, metrics, and correlated logs, pins trusted tool definitions, and detects…

World of Guilds is a leaderless, autonomous multi-agent MMORPG economy. Three AI agents — Ragnar, Lyria, and Kira (powered by NVIDIA Nemotron) — fight, complete quests, and trade on-chain real estate, coordinating entirely through Tashi FoxMQ's Byzantine-fault-tolerant consensus mesh. There is no…

Built a Python-based Geographic Latency Simulator to model remote edge node delay and network hops across distributed microservices. Used OpenTelemetry to instrument HTTP request pipelines and emit trace spans and error metrics directly to SigNoz for visualization.

Talos Observability is a black-box recorder for autonomous browser-testing agents. It makes complex AI test runs fully observable by correlating every API request, queue transition, browser action, LLM call, network failure, token, cost, bug, log and final run outcome in SigNoz through…

CacheGuard is an enterprise observability and active defense middleware designed to secure GenAI prompt caching against Timing Side-Channel attacks. As organizations scale their LLM applications, prompt caching (like vLLM) has become essential for reducing Time-To-First-Token (TTFT) latency and API…

Weft is an autonomous coordination layer that replaces the manual trust infrastructure behind milestone-based capital release — identity, funding, verification, and settlement — with a deterministic, observable agent swarm. Builders stake ETH against project milestones; when a deadline passes…

Whodunnit is an observability and auto-diagnosis layer for multi-agent LLM systems. Traditional monitoring tells you a run was slow or errored. It can't tell you a run succeeded and was wrong. Whodunnit reconstructs each agent run from its OpenTelemetry traces in SigNoz, judges the transcript…

Sentinel MCP Signoz- Signoz Alert based - Agentic system with Unified Observability Dashboard. 1. Agentic application with SREAgent, Data Quality Agent. 2. SREAgent would trigger on Alert (from log/trace/metric) and do RCA. 3. Data Quality Agent would run at certain durations once the Data…

wardn is a version regression detector for your applications. Most teams find out a new release made things worse in one of two ways: a customer complains, or someone happens to be staring at a dashboard at the right moment. Alerts don't catch this either - they fire on fixed thresholds, not on…

AI SRE Agent is an autonomous incident investigator built on SigNoz, OpenTelemetry, FastAPI, and Gemini function-calling. Instead of following a fixed script, it dynamically decides which SigNoz queries to run — slow traces, error logs, error traces — narrowing in on the failing service step by…

Temporal Workflow SLO & Root Cause Correlator is a production-grade platform designed to detect and debug silent failures in complex distributed Temporal workflows. It goes beyond basic monitoring by natively correlating custom OpenTelemetry traces, Temporal SDK metrics, and structured logs into a…

SigNoz Beacon turns live SigNoz alert state into an ambient signal you can't ignore — a browser orb (or Hue / ESP32 LED) that goes green, breathes amber, flashes red, or pulses violet as alerts fire and resolve. A FastAPI bridge accepts SigNoz Alertmanager-style webhooks, maps severity/alertname →…

KathaQuest is an AI lesson studio that turns any textbook chapter or PDF into one coherent, interactive video lesson for children. It plans the pedagogy, writes a scene-by-scene script and storyboard, retrieves relevant real-world footage through VideoDB, adds diagrams and captions, narrates the…

Cadence is telemetry for real-time multimodal agents: a versioned OpenTelemetry semantic convention, an instrumentation SDK, and a Conversation SLO — shipped to SigNoz. Every LLM observability tool assumes request → response. Real-time voice agents hold a persistent duplex stream where neither…

Guru AI Monitor is an autonomous, stateful Discord study agent engineered for rigorous academic preparation, moving beyond stateless LLM wrappers through an embedded SQLite state machine. Track Alignment & Observability Implementation: - Track 01 (AI & Agent Observability): Instruments…

This project demonstrates AI Agent Observability using Python and Streamlit. It includes a simple AI agent that responds to queries and logs inputs, outputs, latency, and errors. These metrics are visualized in a real-time dashboard.

BlackBox AI is an AI observability platform that helps developers trace, debug, replay and optimize AI agent workflows using OpenTelemetry and SigNoz. Modern AI systems often behave like black boxes, making it difficult to understand failures, latency spikes, retries or unexpected outputs. BlackBox…

Meridian is a prompt regression gate for AI agents. It captures a known-good agent trace as a golden baseline, replays the same input after prompt changes, queries SigNoz for trace data, computes a regression verdict, and posts one updateable comment to the GitHub pull request with a link back to…

Punchline Ops is an observable AI workflow for creating traditional memes. A creator enters a familiar situation and receives a rendered WebP meme selected from a curated catalog of 628 classic templates. Behind the simple interface, the server runs deterministic intent extraction, 768-dimension…

OmniSRE is a closed-loop autonomous AI agent that detects incidents, diagnoses root causes, and self-heals infrastructure in under 40 seconds. It wires OpenTelemetry traces into a self-hosted SigNoz stack, uses Gemini 1.5 Flash to programmatically analyze raw ClickHouse error spans, and features a…

Agent K is an autonomous AI SRE that investigates production incidents through SigNoz and is itself fully observable in the same SigNoz instance. A SigNoz alert fires → a webhook wakes the agent → it investigates over the official SigNoz MCP server (traces, logs, metrics, Query Builder v5) → posts…

Preflight checks whether an AI agent was allowed to do what it did. The problem I ran into is that an agent can finish a task, return HTTP 200, throw no errors, and still do something wrong. The trace looks completely healthy. My demo is an SRE agent that handles incidents. The incident ticket has…

TrueSLA finds the customer your dashboard is hiding. Every B2B company signs an SLA: "your service will be up 99.9%, or you get money back." But the dashboard you check that promise on doesn't know your customers exist. It averages everyone into one number. So a small customer can be completely…

AgentOps Studio is an AI agent observability platform that doesn't just monitor an AI agent.it lets the agent explain its own failures. A 5-step research agent (plan --> search --> LLM call --> critique --> finalize) is instrumented end-to-end with OpenTelemetry, exporting traces and…

Ram Store Observability is a production-ready MERN Stack e-commerce application instrumented with OpenTelemetry and SigNoz to provide complete observability. The project demonstrates how to monitor a real-world Node.js and Express backend using distributed tracing, application metrics, structured…

An AI agent that correlates SigNoz observability data with GitHub repositories to identify production issues and suggest code-level solutions using GPT-5. For completely running the project , please download the backend repo as well as frontend repo.

Sella : The Agentic Commerce Economy Sella is an autonomous marketplace infrastructure purpose-built for AI agents to participate as first-class economic actors. Today's AI agents cannot buy data or APIs on their own, they require human operators to manually manage API keys, subscriptions, and…

Agency Immune is a closed-loop self-healing control plane for multi-agent production on SigNoz. It detects six typed agent diseases (tool sepsis, retry storms, token hemorrhage, latency fever, context bloat, telemetry necrosis) from SigNoz alerts and OTLP (gen_ai / tool.exec / agent.name)…

Aegis is an SRE copilot built to close the gap between raw observability data and actionable incident understanding. Modern platforms like SigNoz already store rich traces, logs, metrics, and alerts, but when something breaks, on-call engineers still spend critical minutes manually correlating…

This is an agentic OCR-to-translation pipeline using Node.js/Express, OpenTelemetry, Groq's Qwen 3.6 vision model, and a self-hosted SigNoz instance. The pipeline makes one autonomous decision: after extracting text from an image, it evaluates the model's confidence score and decides whether to…

agentspec

KubeTokenWatch is an AI agent I built for this hackathon that checks a Kubernetes cluster for anything broken and tells you exactly what's wrong and how to fix it. Here's how it works: you give it a namespace, and it scans for failed pods, deployments, jobs, and other objects, not just a fixed list…

AgentGuard: A Behavioral Firewall for AI Agents AI agents don't just talk anymore they act: sending emails, applying discounts, writing to CRMs. That moves the risk from what a model says to what it does, where a single prompt injection or reasoning slip can cause real damage. AgentGuard sits in…

SignalGuard is an autonomous observability drift detector that identifies when an application is still working but its telemetry is incomplete or degraded. It compares current and baseline traces, logs, metrics, and signal correlation against an observability contract, then returns evidence backed…

The problem. When something breaks in production, the alert takes minutes to even fire, because alert rules evaluate on fixed windows — reasonable for paging a human, too slow for anything that could act faster. Once paged, a human has to open five dashboards and manually correlate traces, logs…

TraceGate is a simple safety gate for AI agents before they go live. It runs real-world style test runs of your agent, sends the telemetry to SigNoz with OpenTelemetry, and checks if everything important is visible: main run span, LLM calls, tools, model name, and cost. If something is missing or…

Signoz Studio a small investigation console on top of Signoz. It compares a healthy run to a broken one, explains the failure with evidence from spans, projects what a fix would have saved, and drafts a Signoz alert from the finding.

AegisAI is an AI-powered cybersecurity platform built with FastAPI, Next.js, PostgreSQL, Redis, and Docker, with end-to-end observability powered by OpenTelemetry and SigNoz. The project demonstrates how modern AI applications can be monitored using distributed tracing, metrics, and centralized…

Siggy is an AI-powered incident investigation CLI that sits alongside SigNoz as a sidecar. It learns from every fix, builds a knowledge graph of your infrastructure, and returns explainable recommendations grounded in your operational history.

RecallRoot is a cross-session causal debugger and safety control room for memory-using AI agents. It detects cases where an agent executes a technically successful tool call for an unsafe business reason, traces the decision back to the durable memory that influenced it, and reconstructs the causal…

Rosetta is an OpenTelemetry pipeline layer that gives polyglot AI-agent telemetry one canonical vocabulary. It normalises seven telemetry dialects used by frameworks such as Pydantic AI, CrewAI, LangChain, and Langfuse, calculates model costs, redacts secrets before storage, detects dark and…

A chat assistant app built with Streamlit and Groq's Llama-3.3-70b-versatile model as an LLM, instrumented with OpenTelemetry and visible in a self-hosted SigNoz instance. All the user queries are captured as spans, with the whole request, the LLM API call itself, tokens count with the cost in USD…

Monitors in Black is a hackathon build for the Agents of SigNoz track (AI & Agent Observability). It's a chatbot service dressed up as a noir "city," watched over by a fictional agency whose job is catching anomalies before any human notices them. The real problem behind the theme: an LLM service…

SignalScope is an LLM-orchestrated stock-prediction agent built to answer a question most AI systems can't: *why did the agent decide that?* AI agents rarely fail loudly. They fail by producing a plausible answer from degraded inputs — the news API times out, sentiment silently defaults to neutral…

AgentEye is a small AI agent — powered by Google Gemini with tool-calling — that answers questions using four tools: a calculator, a live weather lookup, a local SigNoz/OpenTelemetry FAQ search, and a clock. The agent itself is intentionally simple; the point of the project is that every single…

Agent Autopsy is an AI support-triage agent instrumented with OpenTelemetry to expose the failure modes that conventional dashboards miss. It ships three injectable failure modes — a slow database shard, a flaky downstream dependency, and prompt-context bloat. The third is the key insight: it…

Agent Black Box is a flight recorder for autonomous AI agents. When an agent fails — retries a dead tool, hallucinates from thin sources, or burns runaway tokens — there's normally no way to reconstruct what happened. Agent Black Box instruments every stage of an agent's execution with…

From Blind Spot to Full Observability: Instrumenting a Live Minecraft Server with SigNoz Running a self-hosted game server means running production infrastructure with zero tooling to back it up when performance degraded, I had no signal, only guesswork. This project treats a Paper Minecraft server…

Migo turns your existing PromQL, LogQL and TraceQL queries into ones SigNoz runs (native builder panels, or compiled ClickHouse SQL for what the builder can't express) and pushes them straight to a live dashboard.

omen is an AI agent that load-tests a code change, finds the regression in live SigNoz traces, and writes the fix, before it reaches production. Point it at a git diff or a plain-language intent. It drives real k6 load against the affected endpoints, correlates the result with server-side telemetry…

SentinelOps is an intelligent observability and incident-response platform built using SigNoz, OpenTelemetry, Node.js, and AI-assisted root-cause analysis. It helps developers monitor distributed applications, automatically detect failures, and quickly understand what caused an incident. The system…

CrisisGrid is aan observability-driven Emergency Operations Intelligence Platform designed to help emergency teams monitor and understand system behavior during simulated crisis scenarios such as floods, fires, road accidents, and notification failures. The platform consists of a React frontend and…

Aegis-Observe is an AI SRE copilot that closes the loop between seeing an incident and fixing it. It runs a continuous diagnostic loop that mines SigNoz telemetry through the MCP server, uses an LLM (Azure OpenAI gpt-5-mini) to select a safe remediation from a fixed toolbelt, and routes the…

RootCause is an observability platform for a hydroponic farm. Most observability watches software; we pointed it at a living system and found a hydroponic reservoir maps almost one-to-one onto the concepts SREs use daily: a nutrient-dosing cycle is a distributed trace, plant vitals (pH, EC, water…

MANDATE is a universal, plug-and-play authority and observability layer for action-taking AI agents. It works with any AI harness and toolset through MCP, REST, or plugins, giving agents bounded access without exposing operational credentials. With SigNoz, it traces every decision from evidence and…

Aro is an AI SRE agent that sits on top of SigNoz and helps teams debug incidents in their AI apps. When latency or errors spike, SigNoz detects it and fires an alert. Aro listens to those alerts, opens an incident, and starts an investigation using SigNoz traces, metrics, and logs. From there, Aro…

This project transforms a self-hosted vLLM inference server into an enterprise-ready AI system. By applying strict SRE principles, it intercepts LLM traffic to track Latency SLAs (Error Budgets), parses token usage for per-team FinOps cost attribution, and runs zero-latency heuristic checks to…

Spring Boot + SigNoz Observability Lab is a self-hosted observability project built around a small e-commerce API. The application is instrumented using the OpenTelemetry Java Agent and manual spans to collect distributed traces, business metrics, and structured logs into a self-hosted SigNoz…

Helios is an AI observability agent that investigates production issues using SigNoz (traces, logs, metrics, k8s events). You ask it questions in natural language (or run tasks like deploy health, RCA, and autoscaling). It queries live telemetry, compares against baselines, and returns evidence…

TraceMind AI is an AI-powered observability platform that helps engineering teams detect, analyze, and resolve production incidents faster. It collects distributed telemetry data using OpenTelemetry and SigNoz, computes key SRE metrics such as latency, error rate, and service health, and uses…

Blindspot is a self-healing AI agent runtime built on SigNoz that catches the failures conventional observability can't see. When a tool silently returns garbage, every trace still shows status OK and normal latency — so the agent emits a custom tool.output_valid quality signal that reveals the…

Root-Cause Attribution of Agents A pipeline-based solution that uses a swap-based ablation engine to correlate LLM agent failures to the underlying reasons (search, tool, generation). FastAPI + Groq API backend and React dashboard frontend were used in its construction. How it operates 1. The…

Production agent failures don't reproduce, so you can't isolate the step that broke or prove a fix before shipping. Rewind streams every step of a multi-agent run into SigNoz as OpenTelemetry spans (I/O, model, tokens, cost), then lets you pull a failed run back, scrub its timeline, and fork one…

I built MarginTrace because most AI observability stops one step too early. A trace can show that every model and tool call completed successfully, while the customer’s actual request still failed. It can also show token usage without explaining whether that spend created value or became waste…

Agent Sherlock is an AI-powered incident investigation platform built on top of SigNoz and OpenTelemetry for autonomous AI systems. It automatically collects distributed traces, reconstructs execution flows, identifies probable root causes using AI-powered reasoning, and generates executive-ready…

SigNoz SLO & Error-Budget Pack is a production-grade, drop-in observability suite for SigNoz. It instruments applications across all three OpenTelemetry signals (Traces, RED Metrics, and correlated Logs) shipping custom dashboards (Service Health, SLO/Error-Budget, Cost-Efficiency), multi-window…

Generative AI agents are operational black boxes. A single infinite loop or poor retrieval window can burn through a startup’s budget overnight, and traditional AI observability tools only show you damage after it has already occurred. AgentLens is an observability-driven SRE copilot for AI agents…

A Telegram weather bot with full SigNoz observability. Every command, API call, Redis lookup, and error is traced end-to-end with OpenTelemetry. Built with aiogram, Redis caching, circuit breaker pattern, and 5 custom metrics.

EvoMind Observability is an observability-first behavioral learning system for AI agents. Built with FastAPI, OpenTelemetry, and SigNoz, it enables end-to-end visibility into how an AI agent evaluates requests, accumulates behavioral evidence, updates confidence, and applies learned guidance over…

RootSpan is a read-only, human-in-the-loop incident correlation system that helps responders find the first broken operation instead of simply following the loudest alert. When an incident occurs, RootSpan compares matched healthy and failing trace cohorts, aligns operations across the service…

Mender, from stack trace to pull request. When an AI agent fails in production, it usually doesn't crash. A tool call throws, the model apologises, and your user gets a confident wrong answer. You find out from a customer. Mender catches it. It reads your agent's traces out of SigNoz, groups…

RingOps: an autonomous on-call agent for the "Agents of SigNoz" hackathon. It detects incidents via SigNoz traces, investigates root cause with the SigNoz MCP server and Groq, then phones the on-call engineer with a voice briefing and lets them act (Slack post, escalate, GitHub issue) via a DTMF…

ArcNet is observability and active defense for AI agent fleets. Every agent is traced end-to-end into self-hosted SigNoz (OpenTelemetry / OpenInference), and every source it ingests carries a trust level via unplug-ai guardrails at four checkpoints, so prompt injection through scraped content is…

Token-efficient AI router agent with full OpenTelemetry observability in SigNoz, plus a terminal SRE sidekick. router-agent classifies tasks (regex) and escalates free solvers → local Ollama self-consistency → at most one Fireworks call. Instrumented end-to-end with OTel spans (task.process…

AgentGlass makes every AI agent run transparent, debuggable, and self-diagnosing — powered by SigNoz Cloud. A CrewAI research agent is fully instrumented with OpenTelemetry using GenAI semantic conventions. Every LLM call, tool call, and agent decision is captured as a SigNoz trace with token…

signoz-init — zero-config observability for AI applications. A Go CLI that inspects any Docker Compose stack, detects each service (FastAPI, Node, Qdrant, …) and its AI libraries, and wires up full OpenTelemetry instrumentation into SigNoz with ONE command — HTTP, database, and LLM/vector-DB spans…

AgentOS AI is a real-time multi-agent AI observability platform that demonstrates how OpenTelemetry and SigNoz can be used to monitor, analyze, and debug AI agent workflows. The system consists of four specialized AI agents—Manager, Researcher, Coder, and Writer—that collaborate to solve user…

TraceMind is an autonomous AI first responder for Site Reliability Engineers, built for the AI & Agent Observability track. When production incidents occur, engineers normally waste precious time manually sifting through dashboards. TraceMind solves this by acting as an autonomous SRE. Instead of…

Faultline is a self-hosted investigation console for AI agent runs. It turns a real OpenAI-compatible mobile-app workflow into an operator-readable causal path, then uses self-hosted SigNoz as the evidence store for traces, logs, metrics, and alerts. Instead of building another observability…

Preflight: catch AI agent regressions before they ship.

ApexVoter is a secure online voting platform that focuses on both security and observability. It provides a complete voting workflow while allowing administrators to monitor the application's health in real time.

AgentLens started as a simple question — if my agent does something wrong in production, will I even know? It's a full agentic RAG system built with LangGraph, Gemini, Pinecone, and Supabase — but the real work was making it observable. Every decision the agent makes is instrumented: LLM calls…

AgentGuardian is an automated LLM observability framework designed to secure and monitor AI Agents. Built using Groq's fast Llama-3.1 engine and OpenTelemetry, it structures deep execution traces to evaluate agent prompts, response latencies, and token anomalies seamlessly via SigNoz patterns.

AI-agent failures are difficult to reproduce because an execution depends on more than application code. Prompts, retrieved documents, tool responses, model decisions, configuration, and external state can all influence the result. ReTrace turns a failed AI-agent execution into a safe, replayable…

ReasonTrail is a flight recorder and safety lab for consequential AI-agent actions. It detects suspicious decisions, reconstructs what happened from SigNoz telemetry, produces an evidence-backed incident explanation, safely replays the execution with a corrected policy, and converts the fix into a…

Sidechalk is a collaborative, AI-assisted living notebook where learners can draw, write, import media, and speak naturally with an assistant that understands the active page. Unlike a conventional chatbot, Sidechalk can propose changes directly on the working surface: handwriting, arrows…

Otto is a self-hosted agentic copilot for SigNoz. It automates the hardest 20% of observability work through four core features: 1. Coverage Audit: Scans live traffic to find telemetry gaps (e.g., services emitting traces but missing logs). 2. Dashboard Migration: Automatically translates Grafana…

Janus is a browser extension that converts between PromQL and SigNoz Query Builder state, in both directions, directly inside SigNoz's UI. Paste or load a PromQL query and convert it to a Query Builder panel, or point it at an existing dashboard panel and convert its Query Builder state back to…

BurnRate is an LLM observability dashboard that tracks AI token usage, estimated costs, latency, and request types. It uses OpenTelemetry to collect these signals and SigNoz to visualize them through traces, metrics, and dashboards. It helps developers quickly identify expensive requests, monitor…

PlanSpan turns Postgres's own query plans into real OpenTelemetry spans in SigNoz. Every plan node, Seq Scan, Hash Join, Sort, becomes a child span under the HTTP request that ran it, plus what-if index fixes, plan regression alerts, lock forensics, and AI diagnosis via SigNoz MCP.

Industrial IoT Anomaly Control is a real-time monitoring agent for industrial sensor data. Traditional factory alarms usually trigger when a sensor crosses a fixed threshold, but they provide little context. They do not explain the possible cause, compare the event with previous incidents, or show…

Hindsight started with a practical problem: SigNoz could show me exactly where an AI agent failed, but I still couldn’t test whether changing one bad input would fix the rest of the run. Each run is recorded in SigNoz as OpenTelemetry traces, correlated payload logs, and metrics for errors…

RestaurantOS AI is an observable, multi-agent restaurant operating system designed to automate day-to-day culinary and logistical workflows. Real-world restaurant management requires tight coordination between forecasting customer traffic, tracking ingredient inventory, reducing food waste, and…

Sentinel - A transparent security proxy for MCP

Agent Black Box gives an AI agent a flight recorder and an on-call copilot, both built on SigNoz. The agent is a small LangGraph support bot that handles refund tickets by looking up orders, querying a policy vector store, and calling tools. Every LLM call, tool call, retry, and RAG lookup is…

I built an AI agent that catches a cascading failure across 4 microservices, figures out which one is actually causing it, fixes it, and checks that the fix worked, all using SigNoz. The setup: Order, Payment, Inventory, and Notification services call each other in a chain, all instrumented with…

BurnRate is a FinOps observability and self-healing cost-control layer for AI agents. OpenTelemetry’s GenAI specification tracks how many tokens an agent consumes, but not how much those tokens actually cost. That means when an autonomous agent gets stuck in a retry loop, uses the wrong model…

SRE Sidekick is an AI-powered observability teammate. It doesn't just read dashboards—it actively validates telemetry by injecting controlled chaos, and autonomously self-heals Kubernetes infrastructure by analyzing metrics and scaling deployments. The entire "thought process" of the AI agent is…

Fuse — An OTel-Native Circuit Breaker for Runaway AI Agents AI agents can enter expensive failure modes without crashing. An agent may repeatedly call an LLM, retry the same tool, circulate between an analyzer and verifier, or continuously expand its context. From an infrastructure perspective, the…

You can't own what you can't observe. Ouroboros is an AI agent that emits full OpenTelemetry telemetry into SigNoz and then reads it back through the SigNoz MCP server to reason about its own latency, cost, and health. The snake eats its own tail.

Most AI agents fail quietly. They complete the task, return a result, throw no errors — and still get it wrong. Traditional monitoring only tells you what happened, not why an agent made a bad call. I built VOID to close that gap. It sits on top of OpenTelemetry and SigNoz, watching agent…

Aegis is a self-healing reliability platform that uses SigNoz telemetry and Codex to investigate incidents, propose bounded remediations, enforce policy or human approval, and verify recovery before declaring success.

CareerPilot AI is an AI-powered career assistant that helps students and job seekers analyze resumes, identify skill gaps, get personalized job recommendations, and practice interviews using AI. The platform uses a React frontend and FastAPI backend with AI-powered services. It provides resume…

RudraOne is an AI agent for emergency dispatch systems. When someone calls 112, the AI answers, listens, triages the emergency, collects the location, calms the caller, and hands off to a human dispatcher with full context. It speaks twelve Indian languages, works in real time, and uses SigNoz to…

An OpenTelemetry instrumentation library and proposed semantic convention for browser agents makes Browserbase and Stagehand workflows observable across a fleet, not one session at a time OpenTelemetry has semantic conventions for HTTP, databases, and LLM calls. There is none for browser agents…

An OpenTelemetry-native engine, built on SigNoz, that detects, classifies, and remediates cloud cost leaks and GenAI quality failures in real time using deterministic cross-signal correlation, not black-box ML.

Receipt proves whether an AI agent actually completed what it claimed to complete.

opentelemetry-instrumentation-crewai is a Python OpenTelemetry instrumentation package for CrewAI, a multi-agent LLM framework that currently ships with no native OTel support (built against a real request from SigNoz's own maintainers, SigNoz/signoz#11663). It emits one root span per crew run, one…

ChronoLens is a closed-loop predictive SRE control plane built entirely on SigNoz. Every reliability tool tells you an outage happened; ChronoLens tells you one is about to, takes a small reversible action to stop it, then proves from your own telemetry that it never landed. The loop is LEARN →…

SRE Sidekick is an AI agent that reads, acts on, and observes SigNoz. Ask it about your services in plain English and it investigates real traces, metrics, and logs via the SigNoz MCP server, returns a root-cause answer, and — with human-in-the-loop approval — creates SigNoz dashboards and alerts…

Gaze is a deterministic verdict engine that watches AI agents through SigNoz traces and issues recomputable scores. Nine pure-Python rules evaluate every span — hallucination, tool abuse, prompt injection, cost explosion, embedding drift. No LLM in the verdict path. Every verdict carries a sha256…

I build a SRE agent that take data from signoz analyse it using Groq and then return problem with solution

FactoryLens is an observability-first manufacturing analytics system that helps explain why production KPIs like OEE change. It instruments the manufacturing data pipeline with OpenTelemetry and sends traces, metrics, and logs to SigNoz, correlating OEE with data-quality signals such as dropped…

Arka is an open-source terminal AI agent. We wired it to SigNoz via OpenTelemetry so every agent step — routing, LLM failover, RAG retrieval, shell tools, self-heal — appears in one trace waterfall. Deploy with arka signoz setup -y, emit demo traces, debug in SigNoz Traces/Services/Logs Explorer…

Arka is an open-source terminal AI agent. We wired it to SigNoz via OpenTelemetry so every agent step — routing, LLM failover, RAG retrieval, shell tools, self-heal — appears in one trace waterfall. Deploy with arka signoz setup -y, emit demo traces, debug in SigNoz Traces/Services/Logs Explorer…

The AI SRE Copilot is an Enterprise-grade DevOps and Site Reliability Engineering (SRE) tool. Its main purpose is to help software engineers and DevOps teams easily monitor their system health, investigate incidents (errors/bugs), and analyze system telemetry (logs, metrics, and traces). This…

Craft My Plate is a role-based restaurant ordering app (admin/manager/user) where the worst failures never crash anything they fail silently. A user hitting an admin-only route gets a clean 403 with no record. An order exceeding stock could go through unnoticed until the kitchen finds out the hard…

AgentOps Center is an AI observability platform for multi-agent systems built around OpenTelemetry, SigNoz, and the Model Context Protocol (MCP). The platform instruments every AI agent interaction, LLM request, tool execution, latency, token usage, and runtime error using OpenTelemetry GenAI…

LoanViz is an AI-powered loan assistant that helps Indian students analyze education loan offers. It uses a LangChain agent with custom tools to calculate EMI, detect hidden fees, assess repayment risk, and provide clear recommendations. The entire agent workflow is instrumented with SigNoz, making…

An autonomous AI Market Research Agent built with the Agent Development Kit (ADK) that streamlines research by gathering and synthesizing information. The agent's LLM calls, tool usage (like Google Search), and token metrics are fully instrumented and monitored using OpenTelemetry and SigNoz to…

SentinelAI is an AI-powered observability platform that helps developers monitor, secure, and debug AI agents in real time. Using OpenTelemetry and SigNoz, it captures traces, logs, metrics, token usage, tool calls, and latency across every stage of an AI workflow. The platform detects prompt…

Taraol is an OpenTelemetry-native observability and runtime protection toolkit for AI-agent applications. Most agent frameworks provide execution logs, but it is still difficult to answer operational questions such as: * Which agent is causing a workflow to become slow or expensive? * Why is one…

GuardrailOps is an open-source, SigNoz-native safety and observability wrapper for LLM chatbots — drop it around any OpenAI, Ollama, or OpenAI-compatible client in 2 lines of code. Most guardrail tools silently block a risky response and move on — nobody knows who triggered it, what they triggered…

SANCTUM is an observable AI firewall. You have an AI agent. People send it messages. Some messages are attacks trying to steal its instructions, jailbreak it, extract secrets, or make it call dangerous tools. Normally you'd never see these until damage was done. SANCTUM sits in front of the agent…

AI Agent with End-to-End Observability (SigNoz) Track 01: AI & Agent Observability AI agents chain multiple LLM calls and tool invocations to answer a single question, but when something goes wrong high latency, runaway token costs, or a bad answer there's no way to see which step caused it. This…

Sherlock is an autonomous incident agent: when a SigNoz alert fires, it investigates traces, logs, and metrics over MCP, returns an evidence-backed root cause, and can page you on Slack — while exporting its own spans so the agent itself is visible in SigNoz.

TelemetryHealth is an intelligent telemetry quality platform built around SigNoz and the OpenTelemetry ecosystem. Instead of monitoring application health, it monitors the health of the telemetry pipeline itself by detecting issues such as high-cardinality attributes, broken traces, missing…

AgentScope is an AI observability platform that helps developers understand what's happening inside their AI applications. As AI systems become more complex with multiple APIs, tools, databases, and language models working together, it becomes difficult to identify why a request is slow, fails, or…

SRE Command Center on SigNoz — 4 SLO dashboards, 9 MCP tools, AI agent (LangGraph + multi-LLM), predictive SLO with burn-rate analysis, auto-remediation poller, canary deploy with auto-rollback, and SLO-as-Code GitOps — all 3 tracks covered.

MERA (Mirror Entity Recursive Agent) is a self-healing AI agent system that uses SigNoz OpenTelemetry + MCP to observe and repair itself in real time. The architecture runs 3 cycles: a Main Agent performs OTel-instrumented code reviews via Ollama + Llama 3.2 3B, the Mirror Agent queries its own…

Agent MedIC is a self-healing AI SRE agent that closes the observability loop. It automatically detects infrastructure incidents via SigNoz alerts, investigates using SigNoz MCP (traces + metrics + logs), diagnoses root cause via local LLM (Ollama), and executes fixes (Docker…

NextOz is a web application that uses AI to answer document-based questions and offers end-to-end functionality. Upload a PDF or TXT file, select a document mode, and ask any questions. The AI reads your document and answers your questions instantly, as if it were thinking out loud on the screen…

GlassPay: Transparent Agentic Finance Built for SigNoz Hackathon Track 01: Build Your Own The Problem AI agents are being given spending authority — autonomous wallets, delegated USDC, Visa cards controlled by LLMs. But when money moves through an opaque pipeline of LLM calls → smart contracts →…

AgentPulse is built on a modern, lightweight stack designed to keep observability as the core — not an afterthought. Backend is Python with FastAPI serving all agent logic and API endpoints. The research agent uses Groq running Llama 3.3 70B for final answers and Llama 3.1 8B Instant for the…

Sidekick watches your Kubernetes cluster through SigNoz and catches problems before you ask — an AI agent that investigates real traces and metrics on its own, and only speaks up with a root cause and a fix in hand, powered by Groq.

MCP Sentinel is a security and observability proxy for AI agent tool calls. Agents now pick their own tools through the Model Context Protocol (MCP), and once you give one real access to your systems, a single prompt injection can make it leak data or run a dangerous command while every individual…

Most engineering teams use Prometheus for metrics and Grafana for dashboards. That's the biggest audience SigNoz wants to bring onto its platform. The biggest barrier isn't the migration itself—it's trust. Teams need to know their dashboards will show the same results after moving to SigNoz. Our…

Trinetra is a debugging platform for multi-agent AI systems, built on SigNoz and OpenTelemetry — it traces every agent, LLM call, and tool call as a distributed system, so when an AI team gives a confidently wrong answer, you can pinpoint exactly which agent caused it and why. We demonstrate this…

ChessWiz / Glass Box Chess Built for the WeMakeDevs x SigNoz "Agents of SigNoz" hackathon (Jul 20–26, 2026). A chess engine application with full-stack observability instrumentation: Chess Engine: C++ engine using negamax with alpha-beta pruning, Zobrist hashing for a transposition table, and…

noz-in is a Go CLI (plus MCP server and agent skill) that migrates a whole Grafana dashboard and its Prometheus alert rules into SigNoz in one command. Every query comes across and renders. The twist: SigNoz's own query engine is the oracle. For each candidate, noz-in fires the native Builder query…

GreenLight is an AI change flight recorder that connects a coding session, immutable commit, CI run, deployment, and the SigNoz telemetry that followed. It produces an evidence receipt showing whether a deployed version regressed the service and whether a later version measurably recovered it.

"See a failure happen. Watch SigNoz catch it. Ask AI why." ChaosCart is a premium, AI-powered e-commerce platform and observability sandbox designed to push SigNoz to its absolute limits. Built specifically for the Agents of SigNoz hackathon, it demonstrates the true power of centralizing your…

The problem Currently Modern AI agent frameworks (LangGraph, Jcode, CrewAI, AutoGen) make it easy to run N agents in parallel. Observability tooling has kept up with single-agent tracing you can follow one agent's reasoning, tool calls, and LLM usage through a flamegraph. What nobody observes is…

RL Observatory Stop flying blind when training AI. RL Observatory brings enterprise-grade observability to reinforcement learning. By wrapping Unity ML-Agents in OpenTelemetry, it streams real-time micro-decisions, reward curves, and structured "Episode DNA" logs directly into a unified SigNoz…

Self Healing SRE Sidekick. Most AI observability demos stop at observe; this one closes the loop to act, safely. A SigNoz alert triggers a local AI agent (qwen2.5:3b on Ollama) that uses self hosted SigNoz, through its MCP server, as a governed closed loop: detect, diagnose, act, verify, roll back…

opentel-mcp is an OpenTelemetry instrumentation library for MCP (Model Context Protocol) servers, built in Node.js and TypeScript, published on npm (currently at v0.4.0). It solves a subtle but common observability gap in MCP-based systems: tool calls can fail silently. A tool can return isError…

We handed the world autonomous AI agents, then shipped them with no instruments. No way to tell when they lie. No way to see them break before they do. The cost is not abstract. A legal copilot citing a court case that does not exist. A triage bot inventing a dosage. A production service quietly…

Eval-Agent · SigNoz Copilot — an SRE copilot for AI agents that lives inside SigNoz. Observability for LLM agents usually stops at "here are the spans." Latency and error rates don't catch a confidently-wrong answer, a dropped citation, or a slow quality regression after a prompt change. Eval-Agent…

Sentinel Copilot is an autonomous SRE agent for "Shadow AI" governance — the problem where teams spin up AI agent containers with root access, exposed ports, and no resource limits, and nobody notices until something breaks. Instead of just another dashboard, we built an agent that watches other…

SigManifest: Making Your Service Catalog Actually Prove Itself Service contracts and architecture diagrams look accurate on day one — then drift silently from reality within weeks. We call this "Manifest Rot," and it's why on-call engineers get paged by dependencies nobody documented and errors…

Reflexon is a self-healing infrastructure system built entirely on SigNoz observability, three services working together: worker-service does real work and can genuinely fail, control-plane holds shared settings plus a permanent incident audit log, and watcher-service is the agent that watches…

Pocket SRE is a mobile-first SRE assistant that brings observability and AI-powered incident response to your phone. It connects with SigNoz to provide real-time metrics, traces, logs, and alerts in a simple mobile interface. Users can chat with AI to investigate incidents, understand failures, and…

an MCP-first AI on-call engineer that investigates incidents by driving the SigNoz MCP server (41 tools) and reports a grounded root-cause diagnosis. Inputs: SigNoz alerts → POST /webhook/alert · Slack @mention & Ask page → POST /investigate · voice → POST /stt (Deepgram) → investigate. Outputs: an…

SigNoz Self-Healing Agent — an autonomous AI SRE that closes the loop most observability tools leave open. It detects an incident using SigNoz, investigates the traces to find the root cause, applies a fix to the running app, then verifies recovery in SigNoz — no human in the loop. In the demo it…

GenAI Observability — Conformance & Quality Layer for SigNoz SigNoz already ships a complete LLM observability product in open source — but a self-hoster can't reach it. Three silent gates block the path: two experimental flags off by default (each redirects/renders empty with no error), and two…

CuraScope is an AI Gateway built for production AI applications that combines AI request processing with end-to-end observability. Every AI request passes through a centralized gateway where it is analyzed for prompt security, routed to the AI provider, stored as a session, instrumented using…

LifeRPG is an observable AI Game Master that transforms daily human habits into a gamified RPG using Telegram, Python, Gemini, and SigNoz. Users text unstructured real-world activities (e.g., "I coded for 3 hours and ate pizza") to a Telegram bot. An AI Game Master parses the text and dynamically…

Canvoz is a prepared-response control plane for known AI-agent failures detected by SigNoz. SigNoz observes our instrumented AI-agent workflows and identifies failures through traces, logs, metrics, dashboards, and alerts. When an alert fires, Canvoz creates a durable incident run, collects a…

The project monitors key observability signals including API latency, distributed traces, HTTP status codes, application logs, MongoDB query performance, JWT authentication events, business metrics, system resource utilization (CPU, memory, network), error tracking, and real-time alerts. Custom…

MergeGuard AI is an AI-powered PR Review Copilot for modern data teams. The agent uses DataHub metadata, lineage, ownership, documentation, and governance context to analyze proposed schema and pipeline changes before deployment. Instead of acting like a chatbot, MergeGuard AI performs a structured…

migobserve is an OpenTelemetry instrumentation library for Go that makes online schema migrations observable in SigNoz. Query-level tracing is already solved — otelsql gives you a span for every SQL statement. What's missing is knowing whether your migration is actually making progress, which phase…

AgentShield is a high performance, out of process security controller that utilizes signoz APM tp automatically detects and terminates runaway agent execution in milliseconds. By keeping active containment independent from the agent's main execution process, AgentShield guarantees that even if a…

Agent Ops Command Center is a real-time observability and governance platform built for multi-agent LLM systems. It monitors live agent execution performance, request latency, operational token costs, and system memory/CPU health. By integrating directly with OpenTelemetry and SigNoz, the system…

SignalForge Command Center is a unified, real-time observability control tower designed for autonomous AI agent architectures. It bridges raw OpenTelemetry (OTLP) traces with human-readable decision intelligence, displaying live agent decision timelines, p95 latencies, simulated token spend, and…

An automated SRE health-monitoring bot that continuously samples system CPU, memory, and disk utilization, evaluates rule-based alert thresholds, and exports structured OpenTelemetry traces with custom attributes directly into SigNoz. Features an embedded HTML web dashboard and JSON telemetry…

LEASH turns SigNoz into an active permission governor for AI agents. When SigNoz detects a downstream error budget breach from OpenTelemetry signals, it POSTs an alert webhook that demotes the agent from T3 (Full Authority) to T1 (Read-Only), intercepting and denying subsequent destructive tool…

ARGUS is a self-hosted, open-source autonomous AI SRE for SigNoz, with one contrarian thesis: everyone builds AI you ask about incidents — but at 2 a.m., an LLM you type questions into is the wrong tool, because a model answering from vibes says something plausible, you chase it, and a confident…

GLASSPANE is browser RUM for self-hosted SigNoz, built on one observation: SigNoz can trace a request from your backend's front door down to the database, but it cannot see the click that opened the door. Track 3 said "observe anything" — and the most valuable "anything" was the one thing SigNoz…

TELELENS is a read-only telemetry cost profiler for self-hosted SigNoz, built on one observation: telemetry bills grow silently and arrive un-itemized. A chatty DEBUG line becomes half your log bill; a user-id label mints a hundred thousand metric series; millions of healthy spans say "everything's…

PharmaTrace is not just a drug interaction assistant; it is an observable AI decision-support workflow where every evidence retrieval step and model synthesis step can be traced, measured, and debugged in SigNoz.

Data teams have distributed tracing for microservices. They don't have it for data pipelines. Hyrax fixes that, it turns dbt runs, Airflow DAGs, and Dagster asset materializations into OpenTelemetry traces and metrics, without touching a single line of pipeline code.

Incident Sentinel is an SRE incident copilot for SigNoz. When an alert fires, a webhook triggers an agent that investigates via the SigNoz MCP server (traces, logs, metrics), posts an evidence-backed Slack report and auto-postmortem, and exports its own OpenTelemetry spans, token usage, and USD…

TraceRoom is a black-box recorder for autonomous financial agents. It records how various AI agents work with the same market information, debate, vote, and go through verification processes. If an agent uses wrong information to back up its position, TraceRoom stops the simulated process and…

QueryPilot turns natural language and PromQL/LogQL into SigNoz Query Builder queries, runs them, and shows live results. It implements SigNoz board issues #11673–#11678 as a working product: NL → BuilderQuery, PromQL/LogQL → BuilderQuery (migration aid), and a one-click gallery of the sponsor…

AgentOps Cockpit is real observability + auto-triage for AI agents on SigNoz. I instrumented a live research agent (Llama 3.2 on Ollama, Wikipedia search, nomic embeddings) with OpenTelemetry GenAI semantic conventions. Every run is one trace; LLM/tool/retrieval steps are child spans with real…

Cortex is an autonomous AI-powered auto-remediation platform built natively for Kubernetes. Instead of relying on static runbooks and manual incident response, Cortex transforms operational knowledge into an intelligent investigation agent that continuously monitors production systems, identifies…

Cortex Cortex is an autonomous AI-powered auto-remediation platform built natively for Kubernetes. It transforms static operational runbooks into an intelligent investigation agent that continuously monitors observability alerts, determines root causes using large language models (LLMs), and…

Coding agents are powerful but opaque. Running Claude Code and OpenCode in separate terminals does not give you a shared plan, isolated parallel work, or visibility into tool calls, stalls, and delivery. **Nova AI** is a CLI multi-agent orchestrator: it discovers agents on your machine, plans…

TraceScene AI is an AI-powered incident investigation platform that combines application telemetry with intelligent analysis to simplify debugging. Instead of manually switching between dashboards, developers can investigate incidents through a unified timeline of services, traces, logs, and…

## Project Description Sentinel is a supervision layer for AI agent fleets. It checks not only whether an agent completed a task successfully, but whether it followed the correct business rules. We built a multi-agent customer-service system with routing, specialist handoffs, structured consent…

A multi‑model AI agent orchestrator (FastAPI + Temporal DAG) that routes each task to a primary LLM with automatic failover to a secondary, made fully observable with a self‑hosted SigNoz. Every step — planning, tool calls, LLM requests, failovers — is traced with OpenTelemetry using the gen_ai…