v0.1.0 — Local Autonomous Memory MVP

OpenMembrain v0.1.0

First release of OpenMembrain — a local-first, private, tool-agnostic memory layer for AI coding tools.

What's included

Core pipeline:

• Session ingestion with secret redaction
• Memory extraction (mock + OpenAI-backed via MemoryExtractor interface)
• Classification, policy/safety/noise filtering
• Deduplication and conflict detection
• Action recommendation (auto_save / ask_user / reject)
• Multi-signal relevance scoring

Storage:

• JSON file-based persistence
• SQLite storage backend (better-sqlite3)
• Memory store, pending candidate queue, audit log, diagnostics log

MCP server:

• 13 tools for memory management, search, approval, export, and diagnostics
• Transport-agnostic ingestion API for adapter integration

Exporters:

• AGENTS.md, CLAUDE.md, copilot-instructions.md, Cursor rules, project-memory.md

Infrastructure:

• TypeScript monorepo (ES2022, ESM, strict mode)
• 34 test files with cross-platform CI (Ubuntu, macOS, Windows)
• Changesets release workflow for future automated versioning

What's next

Phase 4: Build the first thin tool adapter (see roadmap).