AgentTrace — decision recorder for AI coding sessions

Turn AI agent's failed approaches and constraints into reusable team knowledge

Problem

An agent spends an hour exploring a problem, trying different approaches, discovering edge cases, and finally implementing a solution. But everything that didn't make it into the final code—failed approaches, constraints, assumptions, reasons—is lost. Next time, the team and the agent start from scratch.

Solution

Automatically record the full trajectory of an AI coding session: which paths the agent tried, which failed, why they were abandoned, what assumptions were held. Structure this into a searchable decision log and auto-inject relevant history at the start of the next session to avoid repeated exploration.

Users

Developers heavily using AI coding agents, AI-native teams, tech leads

MVP

A local CLI wrapper capturing agent reasoning and tool call traces; extract failed paths, constraints, and assumptions; generate structured decision logs (Markdown/JSON); local full-text search; auto-inject relevant decision summaries before next session

Revenue

Open-source core + paid team version (history sync, cross-project search, access control)

Why now

AI coding sessions are getting longer, yet decision context evaporates after each session—this loss is just being noticed, and capture/summarize tooling doesn't exist