agent-trace

Observability for AI agents. See what your agent did, why it cost that much, and what to fix.

Extensible, low overhead Java Bytecode instrumentation agent for optimizing Java app performance

Long-term memory for AI-coded codebases. A git blame for AI agents — but for the why. MCP server that captures the agent's reasoning live, in context, as each change is made. Local SQLite, zero deps.

The Audited Context Generation (ACG) Protocol prevents AI hallucinations with a dual-layer system. The UGVP layer links every fact to a precise source for verification. The RSVP layer audits the AI's logical reasoning when combining facts. This creates a fully transparent, machine-auditable trail for both source and logical integrity.

Local open-source dev tool to debug, secure, and evaluate LLM agents. Provides static analysis, dynamic security checks, and runtime monitoring - integrates with Cursor and Claude Code.

A curated list of research on LLM-agent traces: evidence tracing, execution provenance, failure attribution, observability, runtime safety, memory provenance, and learning from traces. Companion list of arXiv:2606.04990.

Local-first trace recorder and searchable HTML viewer for AI agent runs.

An MCP server for planning, execution and validation: Structured reasoning (think_*) + execution tracking (ship_*) + roadmap + signals for AI coding agents. Rust · stdio + Streamable HTTP · git-native, Agent-Trace-compatible traces.

🛠️ Enhance RAG systems with ACG to reduce hallucinations in LLM responses by improving accuracy and grounding in reliable external knowledge.

AI 审计网关与责任协议 — 一行 base_url 生成不可篡改的审计轨迹

Alibaba Java诊断利器Arthas

Local-first CLI that turns raw AI agent traces into readable timelines, failure reports, and Mermaid diagrams.

Portable JSON for AI agent sessions. Spec 0.2.0. Converts Claude Code, Grok, Codex, Cursor, OpenCode, and Watchtower 5.x.

面向 DOCX 论文/报告的 AI 格式检测与修改 Agent，支持格式修复、修改报告、Agent Trace、在线预览与下载。

View and debug coding agent transcripts — Claude Code, Codex, and more. Local-first, beautiful, searchable.

AI coding agent with one Python core and three front-ends — headless CLI, Textual TUI, and an Electron desktop. Works with any OpenAI-compatible API, with risk-tiered permissions, event-sourced replayable sessions, and a fail-closed OS-level sandbox.

Agent Trace — visualize an agent's reasoning, parallel tool calls, and results as an interactive DAG inside DeepSeek Harness.

Measure how long AI-authored code actually survives in your repository. Reads Agent Trace records, estimates survival with Kaplan-Meier. Open methodology.