Recall

open source

MCP server that gives Claude persistent memory via Redis-backed semantic search

Updated 2026-03-17

Tool EcosystemsFree· Open-source NPM package; requires Redis instance and OpenAI embeddings APIVisit site

Recall is a TypeScript-based MCP server that solves context loss in Claude by storing and retrieving conversation context as persistent memories. It uses Redis for storage and OpenAI embeddings for semantic search, allowing Claude to maintain project-specific context, coding standards, and architectural decisions across sessions and machines.

Use Cases

·Maintaining consistent project context across multiple Claude sessions
·Storing and retrieving architectural decisions, coding standards, and preferences automatically
·Building knowledge graphs linking related decisions and patterns for complex projects
·Enabling AI coding assistants to apply workspace-specific rules without re-explaining on every conversation
·Creating reusable workflow templates for common development tasks

Key Points

·MCP-native integration, works directly with Claude Desktop, Claude Code, and any MCP client
·Semantic search on 10k+ memories with sub-second latency using Redis + OpenAI embeddings
·Workspace isolation prevents cross-project memory pollution; global memories can be shared intentionally
·27 tools exposed to Claude including memory CRUD, semantic search, versioning, and relationship management
·Lightweight (~189KB) TypeScript bundle; runs locally or with cloud Redis for multi-machine access
·Roadmap includes CI/CD automation, Docker support, and test suite (v1.6.0 in progress)

Recall

Use Cases

Key Points

Tags