# Coding Agent Memory

A local-first memory tool that compiles Codex and Claude Code session history into an auditable, plain-Markdown wiki so coding agents can recall past investigations, decisions and failures instead of repeating them, without a vector database.

This project addresses AI coding assistants re-investigating problems they already solved, by compiling local Codex or Claude Code sessions into a Markdown wiki, then recalling relevant history, engineering decisions and pitfalls on demand through an engineering-memory-loader. Everything is plain Markdown that can be read, edited and version-controlled with git; data stays on the local machine, and nothing is uploaded to a vector database or hosted memory service.

A bounded evidence snapshot of a local session is compiled into a daily wiki entry; a weekly review promotes stable, repeated findings into longer-lived concepts, while ordinary ticket-specific notes are not promoted by default, guarding against the memory reinforcing its own unreviewed guesses. Entries are indexed by keys such as issue IDs, features, repos and tools so related history can be found across sessions, and each daily entry shows a few representative evidence cards that link back to the original session for auditing.

Built with JavaScript, installed by having Codex or Claude Code clone the repository into a project and run its installer, which sets up daily and weekly automation. Apache-2.0 licensed.

Website: https://github.com/miniLV/coding-agent-memory