Home

MCP Local Memory Service

Local-first long-term memory for AI agents: an MCP server (stdio) backed by SQLite + semantic vector search, with a web dashboard and a tree-sitter codebase index. Install once, register it in any MCP-capable AI tool, and your agent keeps durable memory — decisions, code facts, tasks, coding standards, and handoffs — across sessions and projects.

Quick Start

Install the service globally:

npm install -g @vheins/local-memory-mcp

Register the server in your preferred AI tool using the standard mcpServers format (see Getting Started), then start the web dashboard:

npx @vheins/local-memory-mcp dashboard   # open http://localhost:3456

Ask your agent: "What do you remember about this project?" — a list of facts (or a "No memories found yet" message) instead of an error means the integration works.

Quick orientation: Getting Started · Tools Reference · Core Features · MCP Concepts · Troubleshooting

English Documentation

• Getting Started & Setup — install, register, and verify the server
• Tool Reference & Usage Guide — the 17 MCP tools with examples
• MCP Protocol Reference — protocol, resources, prompts
• Core Features — hybrid search, anti-hallucination, knowledge graph, and more
• Hybrid Search — how search ranking works under the hood
• Web Dashboard Guide — command center, tasks, claims, handoffs, knowledge graph
• Troubleshooting Guide — fix common install and runtime issues
• Auto-Start Dashboard — launch the dashboard automatically with your IDE
• Claude Code Integration — connect Anthropic's CLI agent
• Codex (OpenAI) Integration — connect OpenAI's coding agent
• Kiro Integration — connect the Kiro AI IDE

Bahasa Indonesia

Dokumentasi lengkap juga tersedia dalam Bahasa Indonesia:

• Memulai & Pengaturan
• Referensi Alat & Panduan Penggunaan
• Referensi Protokol MCP
• Fitur Inti
• Pencarian Hibrida
• Panduan Dasbor Web
• Panduan Pemecahan Masalah
• Auto-Start Dashboard di IDE
• Integrasi dengan Claude Code
• Integrasi dengan Codex
• Integrasi dengan Kiro

Features

• Codebase Index — index, search, and trace source code symbols
• Long-Term Semantic Memory — durable, hybrid-searchable agent memory
• Task & Project Tracking — structured, state-machine task backlog
• Coding Standards Vault — language & stack scoped coding rules
• Handoff & Claim Coordination — multi-agent handoffs and task claims
• Agentic Session Tools — session briefings, grounded Q&A, repo summaries
• Knowledge Graph — entity–relationship layer over the store

Disclaimer

THE SOFTWARE IS PROVIDED "AS IS", without warranty of any kind. Both developers and agents must collaborate to ensure the quality and accuracy of stored data.