One Slack = Multiple AI Coding Agents
Code anywhere. From anything. Anytime.
Connect AI coding agents (Claude Code, Codex, OpenCode) to Slack. Each channel uses one CLI, different channels can use different ones.
#l0-claude-myapp
|
+-- "build a todo app" --> Claude Code
| "add dark mode" --> Claude Code
| "looks good" --> Claude Code
#l1-codex-backend (different channel = different CLI)
|
+-- "optimize the API" --> Codex
#l0-opencode-utils (another channel)
|
+-- "write unit tests" --> OpenCode
- One Slack workspace, multiple agents
- Multi-round conversations
- Switch agent by renaming channel
- Multiple monitors (5+ windows)
- Claude Code, Codex, OpenCode — all open at once
- Tethered to laptop, can't leave desk
- Every task needs me sitting in front of the computer
#l0-claude-myapp (start new project)
|
+-- "build a react login page"
|
+-- Claude Code: writes code
+-- "looks good, add google auth"
|
+-- Claude Code: adds OAuth
#l1-codex-prod (different channel = different CLI)
|
+-- "deploy to production"
|
+-- Codex: handles deployment
- One Slack = Multiple agents (different channels)
- Code from phone while lying in bed
- Multi-round conversations
- Walk away from desk, let agents work
Work anywhere. From your phone, tablet, or any device with Slack. Your coding agents are always accessible.
The thinking process is streamed to the user.
Add skills to your coding agents (Codex/OpenCode/Claude Code) for enhanced capabilities:
skills/project_workflow— Project workflow automation (build, test, deploy)
⚠️ Important: You must add skills to your coding agents manually. Each agent has its own skill loading mechanism.
Each Slack channel = one project with built-in autonomous level control:
#l0-claude-repo1 → Level 0 (lowest, 80% need user agreement)
#l1-opencode-repo2 → Level 1 (moderate autonomy)
#l2-codex-prod → Level 2 (high autonomy)
- l0: 80% of actions need your approval
- l1: 50% autonomous
- l2: Fully autonomous (trust the agent)
Just rename the channel to switch agent and autonomous level!
 Channel Naming l0-claude-repo1 |
 Migration Rename to switch |
 Status `@BotApp status` to check |
#l1-opencode-repo1 → #l0-claude-repo1
!rate— Check autonomous level and your acceptance rate@OpenClawApp status— Summarize all channels' progress
!plan— One-time plan request!build— One-time build request
Live Demo: [Add your demo URL here]
# Clone and install dependencies
git clone https://github.com/deciding/handclaw.git
cd handclaw
git submodule update --init --recursive
cd openclaw
# Install and build
pnpm install
pnpm ui:build
pnpm build
# Set up Slack connection and install daemon
pnpm handclaw onboard --install-daemonSet these in your Slack config:
{
"requireMention": false,
"groupPolicy": "open",
"streaming": "block"
}requireMention: false— Respond to any message in the channelgroupPolicy: open— Allow any channel to use handclawstreaming: block— Wait for complete response before sending
- Node.js 22+
- pnpm
- Slack workspace
- Install your own coding agents: Claude Code, Codex, and/or OpenCode
⚠️ Note: You must install Claude Code, Codex, or OpenCode separately. handclaw connects to them but doesn't include them.
PRs welcome! This is a fork of OpenClaw — the amazing whale project that makes all this possible.
MIT
handclaw is a personal fork of OpenClaw — the open-source personal AI assistant framework. OpenClaw connects to 15+ messaging channels and supports multiple AI providers.
Check out the main project: github.com/openclaw/openclaw
Try it now →