Show HN: Agentfork – Replay an LLM agent from any step and test fixes

Category: devtools

Tags: llm-agents, replay, debugging, testing, record-per-step, cache

Score: 7.8/10 (Innovation: 7, Technical: 8, Documentation: 9, Utility: 7)

Agentfork is a Python tool for recording, replaying, and forking LLM agent runs at any step. It lets you sweep multiple prompt/model/tool variants with repeats to get pass rates, and save runs as CI tests. That turns flaky debugging anecdotes into measurable experiments with cost estimates, all locally.

Target audience: AI/ML engineers and developers building LLM agents

Repository: https://github.com/bimal1023/agentfork · Python · MIT · 4 stars

View on Hacker News