Open-source AI gateway for local agent clients: publish fusion-style multi-model workflows as OpenAI/Claude-compatible virtual models with traces, tokens, latency, and cost visibility.

MemTrace: Tracing and Attributing Errors in Large Language Model Memory Systems

OpenTelemetry patterns for multi-agent LLM systems. Reference implementation of distributed tracing, metrics, and observability for LangGraph agent orchestration.

OpenAI Agents Tracing | Open-Source Tracing Dashboard including Costs & Usages

Evaluation of Multi-Agent Systems on Cloud Run with the GenAI Client in Vertex AI SDK

Local-first agent debugger with replay, failure memory, smart highlights, and drift detection.

Comprehensive agent analytics suite for AI agents built with the Google Agent Development Kit (ADK) , LangChain or other popular frameworks, powered by BigQuery Agent Analytics plugin and Grafana

Claude Code session capture and analysis. Trace tool calls, detect backtracks, analyze decisions, track edit chains, and measure plan drift. Local-first, zero-dependency observability.

Substructure is an open-source engine that drives your agent loop with JSON webhooks. It keeps sessions durable and streams AG-UI events to your clients. It's language agnostic. You don't even need an SDK.

Open Reflection Protocol — Turn agent failures into regression tests, reusable lessons, and measurable improvements. Built on OpenTelemetry.

Shared context substrate for AI agents. Retrieval that learns what's useful. Runs local or cloud.

Local replay debugger for Browser Use failures with screenshots, model I/O, failed-step timelines, and public-safe HTML exports.

AI coding agent observability and causal tracing for Claude Code, Codex CLI, OpenCode, and Python workflows

Capture and analyze Claude Code sessions locally to track every tool call, decision, and reasoning step without external dependencies.

Homebrew tap for AgentTap - network-level AI agent observability

Observability and governance SDK for AI Agents

ai agent trace, Open-source runtime tracing and diagnostics for AI agent execution flows

Mini PostHog for AI agents: batched TypeScript SDK ingestion into DuckDB (events + traces), an LLM-driven conversational query layer that compiles safe QueryPlans to SQL, and Vantage — a chat-centric trace explorer with live charts, tables, and trace timelines. Production level DuckDB to ClickHouse migration system.

Capture and replay AI agent execution traces. Debug LangChain, OpenAI, and Anthropic agent runs step by step.

A local LLM proxy tool that intercepts and visualizes all OpenAI API calls, providing a real-time web monitoring dashboard.