GitHub Repository

AI agent with multi-agent orchestration, autonomous cognitive systems, and a full management dashboard

46 starsPython

Captain Claw local AI agent, 29 tools, multi-session, DAG orchestration

Name: Captain Claw local AI agent, 29 tools, multi-session, DAG orchestration
Availability: InStock
Author: kstedev

by kstedev·Mar 5, 2026·1 point·1 comment

Visit Project View on HN

AI Analysis

●●SolidBig Brain

Multi-session agents with inter-session context passing, but huge feature list hides unclear priorities.

Strengths

•Session persistence + resume + cross-session context passing is genuinely useful for complex workflows.
•29 tools (shell, files, web, docs, email, calendar, image gen/OCR, etc.) covers breadth—real effort here.
•DAG orchestrator decomposes tasks into parallel multi-agent execution; not trivial.

Weaknesses

•Feature bloat obscures what's actually differentiated—unclear if tools are deeply integrated or just API wrappers.
•No comparison to AutoGen, Crew AI, or LangChain agents, which also do multi-session and DAG orchestration.

Post Description

Captain Claw is a local AI agent runtime I've been building. Install with pip install captain-claw or Docker, point it at any provider (OpenAI, Anthropic, Gemini, Ollama), and get a persistent multi-session agent with a web UI out of the box. The thing I haven't seen elsewhere: sessions are first-class citizens. You can run session #1 on Claude and session #2 on GPT simultaneously, pass context between them, and resume everything exactly where you left off - backed by SQLite. 29 built-in tools the agent picks automatically — shell, files, web search/fetch, PDF/DOCX/XLSX/PPTX extraction, image gen/OCR/vision (DALL-E 3), Gmail (read), Google Calendar, Google Drive, email sending, TTS, todo, contacts, cron, a relational datastore with web dashboard, deep memory via Typesense, playbook distillation, personality profiles, BotPort agent-to-agent routing, and even Termux for Android (camera, GPS, torch, battery). Orchestrator mode decomposes complex tasks into a parallel DAG across separate sessions with real-time progress monitoring. Also ships with an OpenAI-compatible /v1/chat/completions proxy so you can point existing tools at it.

Terminal demo: https://asciinema.org/a/814073

Video: https://www.youtube.com/watch?v=4g_aA_WnEaw

GitHub: https://github.com/kstevica/captain-claw

Happy to answer questions about the architecture or design decisions.