
ai-memory Preserves Context Across Coding Agents
ai-memory is an open-source Rust system that automatically captures coding-agent activity into a versioned Markdown wiki, then exposes shared memory through MCP across Claude Code, Codex, Cursor, OpenCode, and other CLIs. SQLite FTS5 search, handoffs, consolidation, and optional embeddings target the context loss of long-running development work.
ai-memory takes the pragmatic route: durable files for truth, SQLite for retrieval, and agents treated as replaceable clients. Its beta status and limited independent validation are the main risks, but the cross-vendor handoff problem is real and underserved.
- –Automatic capture removes the burden of manually deciding what deserves persistence
- –Markdown remains inspectable and versionable, avoiding dependence on a proprietary vector database
- –MCP support makes project memory portable across competing coding-agent vendors
- –Handoffs can reduce repeated repository exploration when switching agents mid-task
- –Retrieval overhead and stale memories remain unresolved challenges, especially on small or rapidly changing projects
DISCOVERED
1d ago
2026-08-16
PUBLISHED
1d ago
2026-08-16
RELEVANCE