Thread Keeper

@po4erk91

About Thread Keeper

Multi-agent shared brain across Claude Code/Desktop, Codex, Antigravity CLI (agy), Gemini, Copilot, and VS Code. One local SQLite store for threads, notes, verbatim quotes, and a dialectic user model; spawn/broadcast/whisper/inbox multi-agent coordination primitives; autonomous b

#memory #multi-agent #knowledge-graph #persistent-context #cross-cli

Config

Add this server to your MCP-compatible client using the configuration below.

{
  "mcpServers": {
    "thread-keeper": {
      "command": "uvx",
      "args": [
        "--from",
        "threadkeeper[semantic]",
        "python",
        "-m",
        "threadkeeper.server"
      ]
    }
  }
}

Tools

No tools detected

We auto-extract tools from the README. The maintainer can list them under a ## Tools heading to populate this section.

Overview

What is Thread Keeper?

Thread Keeper is a local MCP server that turns multiple parallel agent instances (Claude Code/Desktop, Codex, Antigravity CLI, Gemini legacy, Copilot, and VS Code) into a coordinated multi-agent system instead of isolated chats. It provides cross-session memory, a self-improving skill library, and inter-agent signaling via one shared SQLite store.

How to use Thread Keeper?

Install via pipx, uv, or pip: pipx install 'threadkeeper[semantic]' && thread-keeper-setup. This detects every CLI, registers the MCP server, copies hooks, and writes managed instructions files. Restart your CLI; hook-capable clients auto-inject a brief, while hookless clients (Codex, Antigravity CLI) follow the instructions to manually call brief() or context().

Key features of Thread Keeper

Collective memory: threads, notes, quotes, and dialectic claims shared across all CLIs.
Multi-agent coordination: spawn, broadcast, whisper, inbox, wait, ask, and respond primitives.
Self-improving skill library with autonomous background loops (auto-review, harvester, curator, etc.).
Multi-CLI integration: Claude Code/Desktop, Codex, Antigravity CLI, Gemini legacy, Copilot, VS Code.
Auto-update daemon that checks daily and upgrades via git pull or pip install.
macOS menu-bar app for monitoring loops, cleaning memory, and adjusting settings.

Use cases of Thread Keeper

Share context and memory across different agent CLIs (e.g., Claude and Codex) so skills transfer.
Coordinate parallel agent instances working on the same task, avoiding duplicated work.
Build a persistent, self-improving skill library from past interactions across all CLIs.

FAQ from Thread Keeper

Which CLIs does it integrate with?

Claude Code, Claude Desktop, Codex (CLI and desktop), Antigravity CLI (agy), Gemini legacy, Copilot, and VS Code (every MCP-aware extension via user-level mcp.json).

How does it store data?

All data lives in a local SQLite store. Every connected client shares the same store, threads, user model, and learning loop.

What are the runtime requirements?

Python 3.11+ is required. The recommended install uses pipx; alternatively, uv or a plain venv works. No external database needed—SQLite is bundled.

How does auto-update work?

By default, the MCP server checks once per day. For editable git checkouts, it fast-forwards the tracked branch and reinstalls the editable package. For PyPI/pipx/venv installs, it runs pip install --upgrade. Dirty or diverged git checkouts are skipped.

How does spawning work?

spawn(prompt, slim=True, ...) launches a child Claude session via claude -p subprocess. Slim mode loads only the thread-keeper MCP (~500 MB RSS vs ~1.3 GB). Admission control refuses spawns that would exceed THREADKEEPER_SPAWN_BUDGET_MB (3 GB default).

Frequently asked questions

Which CLIs does it integrate with?

Claude Code, Claude Desktop, Codex (CLI and desktop), Antigravity CLI (`agy`), Gemini legacy, Copilot, and VS Code (every MCP-aware extension via user-level `mcp.json`).

How does it store data?

All data lives in a local SQLite store. Every connected client shares the same store, threads, user model, and learning loop.

What are the runtime requirements?

Python 3.11+ is required. The recommended install uses `pipx`; alternatively, `uv` or a plain venv works. No external database needed—SQLite is bundled.

How does auto-update work?

By default, the MCP server checks once per day. For editable git checkouts, it fast-forwards the tracked branch and reinstalls the editable package. For PyPI/pipx/venv installs, it runs `pip install --upgrade`. Dirty or diverged git checkouts are skipped.

How does spawning work?

`spawn(prompt, slim=True, ...)` launches a child Claude session via `claude -p` subprocess. Slim mode loads only the thread-keeper MCP (~500 MB RSS vs ~1.3 GB). Admission control refuses spawns that would exceed `THREADKEEPER_SPAWN_BUDGET_MB` (3 GB default).

Comments

Basic information

More AI & Agents MCP servers

Hass-MCP

voska

Control and query Home Assistant from Claude and other LLMs — a Model Context Protocol (MCP) server.

307

MCP Server - Remote MacOs Use

baryhuang

The only general AI agent that does NOT requires extra API key, giving you full control on your local and remote MacOs from Claude Desktop App

487

MCP Client for Ollama (ollmcp)

jonigl

Harness the power of local LLMs with this TUI MCP Client for Ollama. Featuring all core MCP primitives (tools, prompts, resources), agent mode, multi-server, model switching, streaming responses, human-in-the-loop, thinking mode, model params config, system prompts, and saved pre

768

bsv.cx

Andy Rowe

Timestamp and verify evidence on-chain. Let your agent prove what it saw and when. bsv.cx gives an AI agent verifiable memory. Timestamp a hash of any output, archive a web page as it looked, or fetch a URL with a recei

Model Context Protocol for Unreal Engine

chongdashu

Enable AI assistant clients like Cursor, Windsurf and Claude Desktop to control Unreal Engine through natural language using the Model Context Protocol (MCP).

2.0K