Recall
Shared long-term memory for AI coding agents
A multi-user knowledge graph with per-person attribution, semantic search, and a force-graph explorer. Gives AI assistants persistent memory across sessions, devices, and projects — so they stop starting from zero every conversation.
Features
Multi-user shared brain
Multiple teammates read and write one shared memory graph. Every memory is attributed to the person who contributed it — the team shares knowledge, not just files.
Semantic search
Vertex AI embeddings with pgvector enable natural-language queries over the entire knowledge graph. Ask 'what decisions did we make about auth?' and get relevant context instantly.
Force-graph explorer
Interactive visualization of the knowledge graph — entities, relationships, and sessions rendered as a navigable force-directed graph with real-time filtering.
MCP protocol native
Built as an MCP server — any MCP-compatible tool (Claude Code, Cursor, etc.) can read from and write to the memory graph. Automatic session ingestion via Stop hooks.
Admin panel
Web-based admin UI for managing users, curating memories, viewing session timelines, and onboarding teammates with generated config snippets.
Architecture
Claude Code session → MCP tools → Cloud Run API → Postgres + pgvector (Cloud SQL) → Vertex AI embeddings. Stop hook auto-ingests session turns. Admin panel at /ui for team management.