AITypeScript★ 300⑂ 11 forksMIT
ATaC-team/ATaC
RECORD AND REPLAY AI AGENT TRAJECTORIES
ATaC records full trajectories of AI agent work and allows them to be replayed for debugging and training. Written in TypeScript, it has already gathered 303★ and supports agent‑infrastructure, MCP, and skills.
// KEY FEATURES
- Records each step of an AI agent: actions, observations and rewards sequentially
- Replays saved trajectories step by step for debugging and verifying behavior
- Exports trajectories to JSON and supports import into MCP environments for further analysis
- Compares two trajectories visually, highlighting differences in actions and outcomes
New repositories every 30 minutes
REDDYX AI scans GitHub 24/7 and ships the best AI/ML/Web3 projects to Telegram.
Join on Telegram