iai-mcp
by CodeAbra·★ 120·Score 49
IAI-MCP is a high-performance memory system for AI coding assistants that implements MCP protocol to provide long-term memory.
Overview
IAI-MCP creates a persistent memory layer for AI coding assistants by capturing every conversation verbatim and organizing it through a three-tier memory system: episodic fragments, semantic summaries, and procedural parameters. It uses LanceDB for efficient storage and bge-small-en-v1.5 for embeddings, with AES-256-GCM encryption for security. The system runs as a Python daemon connected to MCP clients via Unix socket, automatically capturing and recalling relevant context without manual intervention.
Try asking AI
After installing, here are 6 things you can ask your AI assistant:
When to choose this
Choose IAI-MCP when you need long-term memory for your AI coding assistant with verbatim recall, low latency, and complete privacy with local-only storage.
When NOT to choose this
Don't choose IAI-MCP if you're not on macOS, need cross-platform support, or require memory systems that work with non-MCP AI assistants.
Tools this server exposes
6 tools extracted from the READMEcapture-hooksnullInstall, check status, or uninstall hooks to capture conversation data
capture-transcriptnullManually capture a conversation transcript
session-startnullRetrieve relevant memory for the start of a new session
daemonnullManage the iai-mcp daemon process
doctornullRun diagnostic checks on the system
cryptonullInitialize or manage encryption keys
Comparable tools
Installation
Installation
- Clone and install:
git clone https://github.com/CodeAbra/iai-mcp.git
cd iai-mcp
bash scripts/install.sh- Install capture and recall hooks:
iai-mcp capture-hooks install- Connect MCP host (Claude Code):
claude mcp add iai-mcp -- node "$(pwd)/mcp-wrapper/dist/index.js"For Claude Desktop, add to config:
{
"mcpServers": {
"iai-mcp": {
"command": "node",
"args": ["/absolute/path/to/iai-mcp/mcp-wrapper/dist/index.js"]
}
}
}FAQ
- What makes IAI-MCP different from other memory systems?
- IAI-MCP is specifically optimized for AI coding assistants with verbatim recall >99% at 10k entries and p95 latency <100ms. It's a local-only solution with no telemetry and runs entirely on your machine.
- Does IAI-MCP work with LLM assistants other than Claude?
- Yes, any MCP-compatible assistant can use IAI-MCP, including those using Codex CLI. The system exposes memory through the standard MCP protocol.
On Hacker News
Recent discussion from the developer community.
- Story by CodeAbra · 2026-05-07
Compare iai-mcp with
Last updated · Auto-generated from public README + GitHub signals.