projectmem

projectmem gives AI coding assistants a persistent record of a project's debugging history and architectural decisions. It's for developers who want agents to carry useful context between sessions and check past failures before repeating them.

• Project history: Records issues, attempted solutions, fixes and decisions in plain files within your repository; Git hooks capture reverts as failed attempts and fix commits as fixes.
• Pre-commit warnings: Checks staged files against recorded failures, unresolved issues and frequent changes, then identifies approaches that have already failed.
• Agent context: Connects Claude Code, Cursor, Codex and Antigravity through MCP, while Markdown exports provide memory to agents without MCP support; compact summaries and context budgets limit how much history enters a session.
• Cross-project lessons: Carries library-specific gotchas between repositories so an agent can use lessons recorded in earlier projects.
• Decision review: Flags decisions that may be stale after related code changes and preserves earlier decisions when replacements supersede them; a local dashboard lets you inspect the history.

The memory layer runs on your hardware and is free, open source software under the MIT license. It requires no account and has no cloud sync or telemetry. Memory storage and retrieval work offline; an optional update check uses the network. Connected coding assistants receive the context supplied to them, so their model requests depend on the assistant and provider you choose.