selfhost.directory

The index of self-hostable projects — ranked, categorized and served with live updates.
Own your data and stop giving it to big tech.

Captain Claw logo

Self-host Captain Claw

v0.8.3

Multi-agent command center for orchestrating AI agent fleets

BetaOpen SourceMIT
Updated14 hours ago
LanguagePython
Setup difficultyAdvanced
ThemeDark mode
Stars
182
Ports
23080 (captain-claw-web research workspace), 25080 (Flight Deck command center)
Docker
kstevica/captain-claw
Install with Docker
Flight Deck Agent Desktop — multiple AI agents running as a coordinated team with live chat panelStopWatch listening screen with red ring filling toward 110-second limitStopWatch working screen showing live agent activity (google_calendar · list events, 12s · 3 steps)StopWatch reply screen with cyan ring while speaking the agent's responseSimple layout on desktop — chat-first view with agents on left, chat in middle, files on rightSimple layout on phone — responsive chat-first interface

What is Captain Claw?

Captain Claw is a self-hosted platform for managing fleets of AI agents. It centers on Flight Deck, a command center where you launch, watch, and direct multiple specialist agents at once. Seven orchestration modes handle different workflows: parallel execution, AI-designed teams, structured debates, ensemble reasoning, collaborative blackboards, a full coding pipeline with independent review, and a finisher that runs until a judge approves. Each agent gets 51 built-in tools and shares six memory layers. A deterministic Flows language automates multi-step tasks, and a virtual file system lets agents share workspaces. The system works with major LLM APIs or runs fully locally via Ollama. Release 0.8.0 adds multi-user support, Microsoft 365 and Gmail connectors, a smart-glasses HUD, and pocket-sized voice devices. It installs via pip, Docker, Docker Compose, or standalone binaries on macOS, Linux, and Windows.

Should you self-host Captain Claw?

Captain Claw scores 46/100 on selfhost.directory (how scoring works) and ranks #132 of 209 in AI Agents. It has 182 GitHub stars.

It shipped 10 stable releases in the past year. The latest, v0.8.3, came out on 27 October 2026. New versions land roughly every 3 days.

Setup is advanced: plan for several services and some ongoing maintenance. An official Docker image is available. It is MIT-licensed and written mostly in Python.

Its closest alternative on selfhost.directory is Orca (91/100, 108 815 stars). It scores higher overall, so compare the two before you pick. See Captain Claw vs Orca side by side.

Features

  • 31 ready-made specialist archetypes
  • 51 built-in tools
  • Agentic coding pipeline (Code)
  • Deterministic Flows automation
  • Gmail sending with guardsOff by default, requires FD_GMAIL_SEND
  • M5Stack StickS3/StopWatch voice clientsAdded in 0.8.0, requires Soniox API key
  • MCP server for Claude integrationAdded in 0.8.0, requires HTTPS and stable FD_JWT_SECRET
  • Meta Ray-Ban Display HUDAdded in 0.8.0
  • Microsoft 365 integrationAdded in 0.8.0
  • Multi-Agent Orchestration
  • Seven orchestration modes
  • Six-layer shared memory
  • Team deployment with per-user authRequires FD_AUTH_ENABLED=true and related settings
  • Virtual File System (VFS)
  • WhatsApp and Telegram bridges

Captain Claw README

How to install Captain Claw

Install Captain Claw with Docker

Docker

$ docker run -d -p 23080:23080
-v $(pwd)/config.yaml:/app/config.yaml:ro
-v $(pwd)/.env:/app/.env:ro
-v $(pwd)/docker-data:/root/.captain-claw
kstevica/captain-claw:latest

Mount config.yaml and .env read-only and persist data to /root/.captain-claw

How to configure Captain Claw

  • ANTHROPIC_API_KEYAnthropic API key for Claude models

    Anthropic API key for Claude models

  • CLAW_MODEL__MODELgpt-4o-miniModel name to use

    Model name to use

  • CLAW_MODEL__PROVIDERopenaiModel provider (openai, anthropic, gemini, deepseek, ollama, openrouter, chatgpt, antigravity)

    Model provider (openai, anthropic, gemini, deepseek, ollama, openrouter, chatgpt, antigravity)

    values: openai, anthropic, gemini, deepseek, ollama, openrouter, chatgpt, antigravity

  • CLAW_TZEurope/ZagrebUser's local timezone (IANA name)

    User's local timezone (IANA name)

  • FD_AGENT_SHARED_SECRETAgents authenticate to Flight Deck's internal routes with this secret

    Agents authenticate to Flight Deck's internal routes with this secret

  • FD_AGENT_SHARINGoffAllow sharing an agent with other users

    Allow sharing an agent with other users

    values: 1, off

  • FD_ALLOWED_HOSTSExtra host names the deck answers to (comma-separated)

    Extra host names the deck answers to (comma-separated)

  • FD_AUTH_ENABLEDfalseEnable JWT-based multi-user authentication

    Enable JWT-based multi-user authentication

    values: true, false

  • FD_BAT_DISABLEDDeck-wide kill switch for Bat orchestration mode

    Deck-wide kill switch for Bat orchestration mode

    values: 1

  • FD_BAT_SPENDAllow real-money spend in Bat mode

    Allow real-money spend in Bat mode

    values: 1

  • FD_GMAIL_AUTOREPLYDeck kill switch for topic auto-replies

    Deck kill switch for topic auto-replies

    values: off

  • FD_GMAIL_SENDDeck kill switch for agent Gmail sends

    Deck kill switch for agent Gmail sends

    values: off

  • FD_JWT_SECRETSecret key for JWT token signing

    Secret key for JWT token signing

  • FD_LOCKDOWNHosted mode: agent shared secret required even from loopback, host-filesystem routes off

    Hosted mode: agent shared secret required even from loopback, host-filesystem routes off

    values: 1

  • FD_PUBLIC_URLThe deck's public base URL (https://deck.example.com)

    The deck's public base URL (https://deck.example.com)

  • FD_REGISTRATION_OPENReopen self-registration after first admin account

    Reopen self-registration after first admin account

    values: 1

  • FD_SIMPLE_CHAToffKiosk mode: non-admins get only the simple layout

    Kiosk mode: non-admins get only the simple layout

    values: 1, off

  • GEMINI_API_KEYAlternative Google Gemini API key

    Alternative Google Gemini API key

  • GOOGLE_API_KEYGoogle API key for Gemini models

    Google API key for Gemini models

  • OLLAMA_BASE_URLhttp://localhost:11434Ollama base URL for local models

    Ollama base URL for local models

  • OPENAI_API_KEYOpenAI API key for model access

    OpenAI API key for model access

  • TYPESENSE_API_KEYTypesense API key for deep memory

    Typesense API key for deep memory

  • WHATSAPP_REACTIONSonEnable emoji reactions on WhatsApp messages

    Enable emoji reactions on WhatsApp messages

    values: on, off

Captain Claw releases & version history

More in AI Automation

Rate this project

Click to rate — no account needed.

0 comments

Posting as guest
Loading comments…