iai-mcp-memory-server

Configures a local MCP memory server with encrypted semantic recall for AI coding assistants.

7|2|Updated May 16, 2026
One-click install
npx skills add https://github.com/reason-machines/mcp-skills --skill iai-mcp-memory-server-reason-machines
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: iai-mcp-memory-server
Source: https://github.com/reason-machines/mcp-skills/tree/main/skills/iai-mcp-memory-server
Command: npx skills add https://github.com/reason-machines/mcp-skills --skill iai-mcp-memory-server-reason-machines

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires lancedb, sentence-transformers, torch, networkx, igraph.

What problem does it solve? AI coding assistants like Claude Code and Codex CLI lose all context between sessions, forcing users to re-explain prior decisions and discussions. This Skill sets up a local MCP memory server that captures every conversation turn verbatim and automatically recalls relevant context at session start. ## Core Features & Use Cases - Automatic Capture and Recall: Installs hooks that record each conversation turn and inject relevant memories when a new session begins. - Encrypted Semantic Search: Stores memories with AES-256-GCM encryption and LanceDB vector embeddings for sub-100ms similarity queries. - Diagnostics and Maintenance: Provides doctor checks, daemon management, migrations, and benchmark suites to verify recall accuracy and latency. - Use Case: A developer working across multiple Claude Code sessions installs the capture hooks, and the next session automatically receives context about previously discussed error handling patterns without any manual prompting. ## Quick Start Set up the iai-mcp memory server for Claude Code by cloning the repository, running the install script, and installing the capture hooks.

Frequently Asked Questions about iai-mcp-memory-server

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I set up long-term memory for Claude Code?▼

Clone the iai-mcp repository, run scripts/install.sh, then run iai-mcp capture-hooks install to register the capture hooks. Connect the MCP server with claude mcp add pointing to the mcp-wrapper index.js, and verify with iai-mcp doctor.

How to add semantic memory search to an MCP server?▼

iai-mcp stores conversation turns in LanceDB with bge-small-en-v1.5 sentence-transformer embeddings, enabling similarity-based recall. Query memory directly with iai-mcp query followed by your search text, or let the SessionStart hook inject context automatically.

Does iai-mcp work with Codex CLI?▼

Yes, install hooks with iai-mcp capture-hooks install --target codex and register the server in ~/.codex/config.toml under mcp_servers. Use --target all to configure both Claude Code and Codex CLI simultaneously.

Why is no memory recalled at session start?▼

Check that the store contains memories with iai-mcp stats, confirm the daemon is running with iai-mcp daemon status, and test recall manually using iai-mcp session-start --debug. Also inspect the recall log in ~/.iai-mcp/logs/ for errors.

What happens if the iai-mcp encryption key is lost?▼

All stored memories become unrecoverable because data is encrypted with AES-256-GCM and there is no key recovery mechanism. Back up ~/.iai-mcp/.key to a secure location, or run iai-mcp clear --confirm to start fresh with a new key.

What are the platform limitations of iai-mcp?▼

iai-mcp is tested on macOS with Apple Silicon and requires Python 3.11 or 3.12 plus Node 18 or later. Linux and Windows support is planned but not currently available, and Python 3.13 may have torch compatibility issues.