ixchio

ixchio / agent-vcr

Public

Time-travel debugging for AI agents. Record execution, rewind, edit state, and resume without re-running.

pypi.orgprojectai-agent-vcr agent ai debugging langgraph observability
16
2
100% credibility
Found Mar 10, 2026 at 15 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

Agent VCR records AI agent executions step-by-step, enabling users to rewind, inspect, edit states, and resume from failures for efficient debugging.

How It Works

1
🔍 Discover the rewind tool

You learn about a handy tool that records your AI helper's work like a video, so you can rewind mistakes instead of starting over.

2
đź“Ą Add recording to your AI

You simply connect the tool to your AI project with a quick setup, ready to capture every action.

3
▶️ Run and record your task

You give your AI a job to do, and it quietly saves a full movie of every thought and step it takes.

4
⏪ Replay to find the snag

When your AI messes up halfway through, you open the recording and jump back to exactly where it went wrong.

5
✏️ Edit and continue

You tweak the faulty part—like fixing a bad instruction—then hit play to resume right from that spot.

6
📱 Watch or review easily

Use the simple screen viewer to step through, compare changes, or share the fixed run with your team.

🎉 Debug super fast

Your AI works perfectly now, and you saved hours of waiting and money on retries—debugging feels magical!

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 15 to 16 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is agent-vcr?

Agent VCR brings time-travel debugging to Python AI agents, letting you record full executions of multi-step workflows like those in LangGraph or CrewAI. Instead of re-running everything after a failure on step 8, rewind to any point, inspect or edit the state, and resume instantly—saving time and LLM costs. It stores traces as local JSONL files with under 5ms overhead per step, plus a TUI debugger, WebSocket server for live streaming, and CLI tools like `vcr-tui` for playback.

Why is it gaining traction?

Unlike LangSmith or LangFuse, which only log traces, Agent VCR enables actual state edits and forking from mid-execution, mimicking time travel debugging tools from C++ or Java but tailored for agent workflows. One-line integrations hook into frameworks without vendor lock-in, and git-friendly JSONL makes diffs and audits trivial. Developers hook on the instant iteration loop: fix a prompt glitch and resume downstream without restarting.

Who should use this?

AI engineers debugging LangGraph graphs or CrewAI crews where failures hide deep in chains. Team leads auditing agent decisions via side-by-side run comparisons. Prompt researchers forking executions to test variations without full replays.

Verdict

Grab it if you're knee-deep in agent debugging—early promise with solid docs, benchmarks, and 100% test coverage outweighs the 13 stars and 1.0% credibility score. Still beta, so pair with production observability until maturity grows.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.