AliceLJY/recallnest

Overview

A shared memory layer for Claude Code, Codex, and Gemini CLI. It provides hybrid retrieval using vector, BM25, and knowledge graph methods, with session continuity and 41 MCP tools. The tool is local-first and stores data in LanceDB.

Best for

Best for
Developers using multiple AI CLI tools who need a unified, local memory layer for session continuity

Use cases

• Persisting and retrieving conversation context across Claude Code, Codex, and Gemini CLI sessions
• Enabling hybrid search (vector, BM25, knowledge graph) over previous interactions
• Accessing 41 MCP tools to extend CLI capabilities with shared memory

How to use

Install

npx recallnest --help          # run directly

Tools exposed

• workflow_observe
• workflow_health
• workflow_evidence
• store_memory
• store_workflow_pattern
• store_case
• promote_memory
• promote_scan
• list_conflicts
• audit_conflicts
• escalate_conflicts
• resolve_conflict
• checkpoint_session
• latest_checkpoint
• resume_context
• search_memory
• explain_memory
• distill_memory
• brief_memory
• pin_memory

Tested with

Claude Code, ChatGPT

Notes

A shared memory layer for Claude Code, Codex, and Gemini CLI. It provides hybrid retrieval using vector, BM25, and knowledge graph methods, with session continuity and 41 MCP tools. The tool is local-first and stores data in LanceDB.

14 stars on GitHub. Last updated 2026-05-31. Licensed MIT.

Use cases

• Persisting and retrieving conversation context across Claude Code, Codex, and Gemini CLI sessions
• Enabling hybrid search (vector, BM25, knowledge graph) over previous interactions
• Accessing 41 MCP tools to extend CLI capabilities with shared memory

Pros

• Local-first architecture ensures data privacy and offline capability
• Hybrid retrieval combines multiple search strategies for better results
• Provides session continuity across three popular AI CLI tools

Cons

• Limited to Claude Code, Codex, and Gemini CLI (no support for other CLIs)
• Small community size (14 stars) may mean less ongoing development and support
• Requires setting up and maintaining a LanceDB instance locally

Indexed from awesome-mcp-servers-punkpeye and enriched against its public facts.