Desktop app · Web console · Service CLI — three ways to run OpenClaw without editing config files.
📦 Releases · 💬 Discussions · 🐛 Issues · Discord | English · 中文 · 日本語
| OpenClaw CLI alone | ClawMaster | |
|---|---|---|
| Initial setup | Hand-edit ~/.openclaw/openclaw.json |
Guided wizard |
| Provider & model config | Edit JSON, restart | Form UI with live validation |
| Channel setup | Read docs, edit config | Step-by-step guides per platform |
| Observability | None built-in | ClawProbe dashboard (cost, tokens, health) |
| Memory management | powermem CLI |
Management UI |
| Multiple profiles | Manual file juggling | Profile switcher |
| Desktop app | No | Yes — ships as .dmg / .msi / .AppImage |
| Self-hosted web console | No | Yes — Express, runs anywhere Node.js runs |
"I manage OpenClaw for my team." One place to configure channels, rotate API keys, and monitor token spend — no SSH, no JSON editing.
"I'm building agents with LangChain." Quick observability into context usage, memory snapshots, and cost-per-session without writing monitoring code.
"I'm setting up OpenClaw for the first time." The setup wizard walks you through provider, model, gateway, and channel in one flow. No docs required to reach a working state.
- Setup and profiles — Detect OpenClaw, install missing pieces, create or switch profiles, bootstrap a local environment.
- Models and providers — Configure OpenAI-compatible and provider-specific endpoints, validate API keys, set runtime defaults.
- Gateway and channels — Bring up the gateway, follow guided setup for Feishu, WeChat, Discord, Slack, Telegram, and WhatsApp.
- Plugins, skills, and MCP — Enable or disable capabilities, install curated items, add MCP servers, import MCP definitions.
- Sessions, memory, and observability — Inspect sessions, manage memory backends, track token usage and estimated spend.
Option 1: Desktop installer
Download the latest installer from GitHub Releases.
| Platform | Format |
|---|---|
| Linux x64 | .deb, .rpm, .AppImage |
| macOS Intel | .dmg |
| macOS Apple Silicon | .dmg |
| Windows x64 | .msi, .exe |
[!NOTE] CI also uploads per-platform artifacts for every push to
main(7-day retention) if you need an unreleased build.
Option 2: Run from source
git clone https://github.com/openmaster-ai/clawmaster.git
cd clawmaster
npm install
npm run dev:web # web app + backend
npm run tauri:dev # desktop appRequirements: Node.js 20+. For Tauri desktop builds, also Rust — see tauri.app/start/prerequisites.
Option 3: Service CLI
npm i -g clawmaster
clawmaster doctor
clawmaster serve --daemon
clawmaster statusDefault service URL: http://127.0.0.1:3001. clawmaster serve prints a service token — enter it in the browser UI when prompted.
Common flags:
clawmaster serve --host 127.0.0.1 --port 3001 --daemon
clawmaster serve --host 127.0.0.1 --port 3001 --token your-own-token
clawmaster stop
clawmaster doctor- Launch ClawMaster.
- Choose an existing OpenClaw profile or create a new one.
- Connect at least one model provider and set a default model.
- Enable gateway or observability if you need runtime inspection.
- Add channels, plugins, skills, or MCP servers as needed.
Six capabilities — tracked as labeled issues:
| Capability | Status | What it covers |
|---|---|---|
| Setup | Released | Wizard, 16 providers, 6 channel types, profile management |
| Observe | Released | ClawProbe integration, cost / token / health dashboard |
| Save | In progress | PowerMem UI, seekdb integration, token-reduction workflows |
| Apply | Planned | Photo OCR, invoice processing, flashcard tools |
| Build | Planned | Conversational agent builder (LangChain DeepAgents) |
| Guard | Planned | Key encryption, spend limits, RBAC |
Browse label:roadmap to pick up an item. Leave a comment before starting — core contributors who land roadmap features can claim model credits from the OpenClaw team.
- 2026-04-13 🏗️ Contribution workflow tightened with issue forms, a stronger PR template, PR description validation, and architecture boundary tests.
npm install
npm run dev:web # frontend + backend
npm run dev # frontend only (port 3000)
npm run dev:backend # backend only (port 3001)
npm run tauri:dev # desktop appTesting and CI
npm test # unit tests (Vitest)
npm run build # type check + production build
npm run test:desktop # desktop smoke (macOS: real Tauri build; Linux/Win: WebDriver)[!TIP] Run
npm test && npm run buildbefore opening a PR — the same steps run in CI.
CI covers: TypeScript check, unit tests, backend integration smoke, web smoke, desktop smoke, and multi-platform Tauri builds.
Project layout
clawmaster/
├── packages/web/ React + Vite frontend
├── packages/backend/ Express backend for web mode
├── src-tauri/ Tauri desktop host
├── tests/ui/ YAML-based manual UI flow specs
└── bin/clawmaster.mjs CLI entry point
Runtime model: Desktop — React calls Tauri commands via invoke(); Web — React proxies /api to Express.
Using an AI coding agent? Point it at AGENTS.md first — it covers the full contribution workflow, module patterns, and hard rules in agent-readable form.
See CONTRIBUTING.md for setup, testing requirements, dependency policy, commit convention, and PR checklist.
Important
PRs must pass npm test locally before opening. No screenshots, test logs, or generated files in commits. Node.js is the only permitted runtime — no new language dependencies.
Community: GitHub Discussions · Discord · Feishu
Acknowledgments
| Project | Role |
|---|---|
| OpenClaw | Core runtime and configuration model |
| ClawProbe | Observability daemon |
| ClawHub | Skill registry |
| PowerMem | Memory backend |
| Tauri | Desktop app framework |
| React | Frontend UI |
| Vite | Frontend toolchain |
| Playwright | Browser automation and smoke testing |
MIT. See LICENSE.