agent-recall - Persistent memory for AI agents
Stores, retrieves, and compounds agent memory across sessions using local Markdown, with optional semantic search through Supabase.
Tags
Updated: 2026-10-08Capabilities
Typical Inputs
Typical Outputs
What this skill does
- Load prior session context
- Save session memories
- Search stored memories
- Track corrections and decisions
- Consolidate session knowledge
- Merge same-day journal entries
- Match cross-project insights
- Generate semantic memory names
Inputs
- Project name
- Session context
- Memory content
- Search query
- Session summary
- Session insights
- Session trajectory
- Goal and assumptions
- Human corrections
- Feedback ratings
Outputs
- Loaded context briefing
- Saved memory records
- Ranked recall results
- Correction warnings
- Decision trail records
- Daily journal entries
- Updated awareness insights
- Updated palace rooms
- Session save summary
Requirements
- MCP-compatible agent
- Running MCP server
- npx and Node.js environment
- Read-write access to ~/.agent-recall/
- Optional Supabase configuration for semantic search