Skip to content

alibaba/hiclaw

Repository files navigation

HiClaw

English | 中文

DeepWiki Discord

HiClaw is an open-source Collaborative Multi-Agent OS for transparent, human-in-the-loop task coordination via Matrix rooms.

Built with a Manager-Workers architecture, HiClaw lets you coordinate multiple Worker Agents through a Manager Agent to complete complex tasks — all conversations are visible in Matrix rooms, and you can intervene at any time.

Think of it as your AI team in a chat room: tell the Manager what you need, it spins up Workers, and you watch everything happen in real-time.

Key Features

  • 🧬 Manager-Workers Architecture: Eliminates the need for human oversight of individual Worker Claws by enabling Agents to manage other Agents.

  • 🦞 Customizable Agents: Each Agent supports flexible configurations including OpenClaw, Copaw, NanoClaw, ZeroClaw, and enterprise-built Agents—scaling from individual "shrimp farming" to full-scale "shrimp farm" operations..

  • 📦 MinIO Shared File System: Introduces a shared file system for inter-Agent information exchange, significantly reducing token consumption in multi-Agent collaboration scenarios.

  • 🔐 Higress AI Gateway: Centralizes traffic management and mitigates credential-related risks, alleviating user concerns about security vulnerabilities in the native Lobster framework.

  • ☎️ Element IM Client + Tuwunel IM Server (both Matrix protocol-based): Eliminating DingTalk/Lark integration overhead and enterprise approval workflows. Enables rapid user onboarding to experience the "delight" of model services within an IM environment, while maintaining compatibility with native OpenClaw IM integration.

News

  • 2026-03-14: HiClaw 1.0.6 — enterprise-grade MCP Server management, zero credential exposure. Blog
  • 2026-03-10: HiClaw 1.0.4 — CoPaw Worker support, 80% less memory. Blog
  • 2026-03-04: HiClaw open sourced. Announcement

Why HiClaw

  • Enterprise-Grade Security: Worker Agents operate with consumer tokens only. Real credentials (API keys, GitHub PATs) stay in the gateway — Workers can't see them, and neither can attackers.

  • Fully Private: Matrix is a decentralized, open protocol. Host it yourself, federate with others if you want. No vendor lock-in, no data harvesting.

  • Human-in-the-Loop by Default: Every Matrix room includes you, the Manager, and the relevant Workers. Watch everything. Jump in anytime. No black boxes.

  • Zero Configuration IM: Built-in Matrix server means no bot applications, no API approvals, no waiting. Just open Element Web and start chatting.

  • One Command Setup: curl | bash and you're done — AI gateway, Matrix server, file storage, web client, and Manager Agent.

  • Skills Ecosystem: Workers pull from skills.sh (80,000+ community skills) on demand. Safe because Workers can't access real credentials.

Quick Start

Prerequisites: Docker Desktop (Windows/macOS) or Docker Engine (Linux).

Resources: 2 CPU cores + 4 GB RAM minimum. For multiple Workers, 4 cores + 8 GB recommended.

Install

macOS / Linux:

bash <(curl -sSL https://higress.ai/hiclaw/install.sh)

Windows (PowerShell 5+):

Set-ExecutionPolicy Bypass -Scope Process -Force; $wc=New-Object Net.WebClient; $wc.Encoding=[Text.Encoding]::UTF8; iex $wc.DownloadString('https://higress.ai/hiclaw/install.ps1')

The installer walks you through:

  1. Choose your LLM provider (OpenAI-compatible APIs supported)
  2. Enter your API key
  3. Select network mode (local-only or external access)
  4. Wait for setup to complete

Access

Open http://127.0.0.1:18088 in your browser and log in to Element Web. The Manager will greet you and explain how to create your first Worker.

Mobile: Use any Matrix client (Element, FluffyChat) and connect to your server address.

That's it. No bot applications. No external services. Your AI team runs entirely on your machine.

Upgrade

# Upgrade to latest (preserves all data)
bash <(curl -sSL https://higress.ai/hiclaw/install.sh)

# Upgrade to specific version
HICLAW_VERSION=v1.0.5 bash <(curl -sSL https://higress.ai/hiclaw/install.sh)

How It Works

Manager as Your AI Chief of Staff

You: Create a Worker named alice for frontend development

Manager: Done. Worker alice is ready.
         Room: Worker: Alice
         Tell alice what to build.

You: @alice implement a login page with React

Alice: On it... [a few minutes later]
       Done. PR submitted: https://github.com/xxx/pull/1

    

① Manager creates a Worker and assigns tasks                    ② You can also direct Workers directly in the room

Security Model

Worker (consumer token only)
    → Higress AI Gateway (holds real API keys, GitHub PAT)
        → LLM API / GitHub API / MCP Servers

Workers see only their consumer token. The gateway handles all real credentials. The Manager knows what Workers are doing but never touches the actual keys.

Human in the Loop

Every Matrix Room includes you, the Manager, and relevant Workers:

You: @bob wait, change the password rule to minimum 8 chars
Bob: Got it, updated.
Alice: Frontend validation updated too.

No hidden agent-to-agent calls. Everything is visible and intervenable.

Architecture

┌─────────────────────────────────────────────┐
│         hiclaw-manager-agent                │
│  Higress │ Tuwunel │ MinIO │ Element Web    │
│  Manager Agent (OpenClaw)                   │
└──────────────────┬──────────────────────────┘
                   │ Matrix + HTTP Files
┌──────────────────┴──────┐  ┌────────────────┐
│  hiclaw-worker-agent    │  │  hiclaw-worker │
│  Worker Alice (OpenClaw)│  │  Worker Bob    │
└─────────────────────────┘  └────────────────┘
Component Role
Higress AI Gateway LLM proxy, MCP Server hosting, credential management
Tuwunel (Matrix) Self-hosted IM server for all Agent + Human communication
Element Web Browser client, zero setup
MinIO Centralized file storage, Workers are stateless
OpenClaw Agent runtime with Matrix plugin and skills

HiClaw vs OpenClaw Native

OpenClaw Native HiClaw
Deployment Single process Distributed containers
Agent creation Manual config + restart Conversational
Credentials Each agent holds real keys Workers only hold consumer tokens
Human visibility Optional Built-in (Matrix Rooms)
Mobile access Depends on channel setup Any Matrix client, zero config
Monitoring None Manager heartbeat, visible in Room

Roadmap

✅ Released

  • CoPaw — Lightweight agent runtime Released in 1.0.4: ~150MB memory usage (vs ~500MB for OpenClaw), plus local host mode for browser automation.
  • Universal MCP Service Support — MCP server integration Released in 1.0.6: Any MCP server can be safely exposed to Workers through the gateway. Workers access tools using only Higress-issued tokens; real credentials never leave the gateway.

In Progress

Lightweight Worker Runtimes

  • ZeroClaw — Rust-based ultra-lightweight runtime, 3.4MB binary, <10ms cold start.
  • NanoClaw — Minimal OpenClaw alternative, <4000 LOC, container-based isolation.

Goal: Reduce per-Worker memory from ~500MB to <100MB.

Planned

Team Management Center

A built-in dashboard for observing and controlling your Agent Teams — real-time observation, active interruption, task timeline, resource monitoring.


Documentation

docs/quickstart.md Step-by-step guide
docs/architecture.md System architecture deep dive
docs/manager-guide.md Manager configuration
docs/worker-guide.md Worker deployment
docs/development.md Contributing and local dev

Troubleshooting

docker exec -it hiclaw-manager cat /var/log/hiclaw/manager-agent.log

See docs/zh-cn/faq.md for common issues.

Build & Test

make build          # Build all images
make test           # Build + run all integration tests
make test-quick     # Smoke test only

Other Commands

make replay TASK="Create a Worker named alice for frontend development"
make uninstall
make help

Community

License

Apache License 2.0

About

An open-source Collaborative Multi-Agent OS for transparent, human-in-the-loop task coordination via Matrix rooms.

Topics

Resources

License

Stars

Watchers

Forks

Packages

 
 
 

Contributors