Nexu Desktop
A local desktop client that brings an OpenClaw Agent into WeChat, Feishu, Slack, and Discord.
- Source repo
- nexu-io/nexu
- Stars
- ★ 3.3k
- Last updated
- 6mo ago
- License
- MIT
- Primary language
- TypeScript
- FA score
- 42/100 · Major gaps
At a glance
- How it runs
- Works with
- Platform-specificCodex · Claude Code
- Cost
- Free software; you pay for model usage
- Setup effort
- Low · running in minutes
- You'll need
- Typical use
- A solo cross-border ecommerce operator uses Feishu to send product specifications to the Agent for multilingual listings, competitor-price research, and marketing assets.
- Not a fit if
- Linux desktop users: only macOS 12+ and Windows 10+ are supported
- Teams unwilling to depend on OpenClaw and the WeChat 8.0.7 plugin
- Teams needing full production deployment and ops documentation
- Source review
- 42/100 · Major gaps 1 safety controls not found
What does this agent do, and when should you use it?
Nexu is an open-source desktop client for connecting an OpenClaw Agent to WeChat, Feishu, Slack, Discord, and other IM channels. The repository includes a web frontend, an Electron desktop client, a controller service, and shared libraries; its full local development stack is started with pnpm. The installed product is operated through a desktop GUI rather than a command line, and connected IM channels let users chat with the Agent from a phone. The README describes Nexu as local-first and states that business data stays on the user's machine. It supports a nexu account for supported models or user-provided API keys, while MiniMax, OpenAI Codex, and GLM (Z.AI Coding Plan) support OAuth login.
A user downloads and opens the desktop client from nexu.io or GitHub Releases, then signs in to a nexu account or adds an API key. The client runs an OpenClaw Agent and offers built-in WeChat, Feishu, Slack, and Discord connections; the documented WeChat path uses the WeChat 8.0.7 OpenClaw plugin, where the user clicks connect and scans with WeChat. Once connected, the user can chat with the Agent in the linked IM channel, including from a phone. For development, pnpm dev start launches the local stack containing apps/web, apps/desktop, and apps/controller, while pnpm build creates a production build.
- A solo cross-border ecommerce operator uses Feishu to send product specifications to the Agent for multilingual listings, competitor-price research, and marketing assets.
- A content creator asks the Agent in Slack for topic ideas and uses it to prepare drafts and engagement work for multiple publishing platforms.
- An indie developer pastes an error or stack trace into Discord for Agent-assisted code analysis, documentation work, or PR-description drafting.
- A legal, finance, or consulting professional forwards contract material in Feishu for risk summaries, flagged clauses, and revision suggestions.
- A small retailer uses an Agent in Feishu to assist with inventory questions, order follow-up, customer replies, and promotional copy.
How do you install or deploy this agent?
Use macOS 12+ on Apple Silicon or Intel, or Windows 10+, with approximately 500 MB of storage. Download the platform build from https://nexu.io or https://github.com/nexu-io/nexu/releases and open it. On first launch, sign in with a nexu account for supported models or add your own API key; the supplied material does not document the provider-specific API-key fields, files, or commands.
How do you use this agent?
In the desktop GUI, connect WeChat, Feishu, Slack, or Discord. For WeChat, use the WeChat 8.0.7 OpenClaw plugin: click connect in Nexu and scan with WeChat. Then send messages to the OpenClaw Agent through the connected chat channel, which can also be used from a phone. For local development, install Node.js 22+ and pnpm 10+, then run pnpm dev start; use pnpm dev stop to stop the stack and pnpm build for a production build.
What are this agent's strengths and limitations?
- Brings WeChat, Feishu, Slack, and Discord into one desktop client instead of requiring separate channel integrations.
- Provides a GUI-based install and connection flow; the README explicitly says users can start without a CLI or environment variables.
- Supports a nexu account, BYOK, and OAuth login for MiniMax, OpenAI Codex, and GLM, giving users multiple credential paths.
- Describes itself as local-first and says business data remains on the machine, while the MIT license permits audit and derivative work.
- The documented desktop support is limited to macOS 12+ and Windows 10+; Linux support is not described.
- Its core use depends on OpenClaw, and channel connectivity remains subject to the relevant platform's plugin or setup requirements.
- The documented WeChat path specifically requires the WeChat 8.0.7 OpenClaw plugin; compatibility with other versions or connection paths is not stated.
- The README does not provide detailed channel permissions, troubleshooting procedures, API-key field definitions, or a production deployment guide for the controller service.
How does this agent compare with similar options?
The README compares Nexu with official OpenClaw, positioning Nexu as a GUI route to ready-made channel connections, model switching, and a local-first data path; the comparison table characterizes official OpenClaw as requiring manual configuration and DIY channel integration.
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| Nexu Desktop This agent | 42 · Major gaps | Desktop appFree + model costs | ★ 3.3k | 6mo ago | TypeScript | Codex · Claude Code |
| Open Cowork | 65 · Some gaps | Desktop appFree + model costs | ★ 2.3k | 17d ago | TypeScript | Claude Code · OpenAI API · Claude API |
| OpenCrew Team OS | 48 · Major gaps | CLIFree + model costs | ★ 496 | 7mo ago | Shell | — |
| Inno Agent | 60 · Some gaps | Desktop appFree + model costs | ★ 1.4k | 1d ago | TypeScript | OpenAI API · Claude API |
How does FollowAgents rate this agent?
Why each dimension lost points
Evidence shows: README claims local-first and data stays on device, but no implementation details; telemetry (PostHog) and cloud identity features exist, but no explicit user consent mechanism; many third-party dependencies, but no security audit; SECURITY.md exists but no vulnerability fix records. Deductions: lack of evidence for least privilege, user confirmation, data flow transparency, sensitive data handling; external effects (e.g., IM integrations) not scoped; no rollback mechanism; source attribution relies only on MIT license and unverified publisher.
Evidence shows: code structure is clear, tests cover some functionality, but no full test results; many third-party dependencies, but no availability guarantees; error handling appears in tests, but no user-visible error messages documented. Deductions: dependency availability unverified; failure messages not documented.
Evidence shows: README clearly defines target users (individuals and small teams) and scenarios (e-commerce, creators, etc.), but no capability boundaries; trigger mechanisms (e.g., IM integrations) not detailed; environment requirements (macOS, Windows) clear. Deductions: capability boundaries and trigger precision insufficiently documented.
Evidence shows: README is well-structured, provides installation instructions, FAQ, community links; MIT license present; but no version changelog and known limitations; naming stability unverified. Deductions: missing version changelog and known limitations; maintenance responsibility unclear.
Evidence shows: README claims value (e.g., time savings), but no actual output examples or user feedback; marginal value comparison table exists, but no cost-benefit analysis. Deductions: output usability unverified; cost-benefit not quantified.
Evidence shows: README claims (e.g., local-first) lack implementation evidence; tests exist but no coverage report; no separation of facts and inferences. Deductions: claim traceability insufficient; cross-source corroboration insufficient.
- Not found in source: rollback or recovery pathBack up first, or work on a git branch or snapshot, so its changes can be undone.
- Publisher identity is unverified; proceed with caution.
- Telemetry (PostHog) and cloud identity features may involve data transmission; ensure user consent mechanisms are in place.
- Many third-party dependencies; recommend security audit.
- Missing version changelog and known limitations; be cautious before upgrading.
Recent releases
Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.
-
Release notes excerpt
🎉 Highlights - 🪟 Windows support is here — Nexu now officially supports Windows as a first-class desktop platform. Install, upgrade, local runtime, tray behavior, and SkillHub extraction all received a full stabilization pass. (#949, #986, #988, #1021, #1030, #1073) Thanks @PerishCode, @alchemistklk, @mrcfps, @lefarcen, and @Siri-Ray. …View on GitHub ↗ -
Release notes excerpt
🎉 Highlights - 🏆 Desktop Rewards Center — Earn credits by starring the repo, sharing assets, and inviting friends. Track your balance and claim rewards directly from the desktop app. (#789) Thanks @anthhub @nettee . - 💬 Three new managed channels: QQ, WeCom, and DingTalk — Connect QQ, WeCom, and DingTalk directly from the dashboard as a bundled new channel set. (#775, #895) Thanks @Siri-Ray and @qiongyu1999. …View on GitHub ↗ -
Release notes excerpt
🎉 Highlights - 🎬 Seedance 2.0 video generation — Generate 15-20 second AI videos right from your bot conversations. Star the repo and join the community group to get free credits. (#749, #751, #764) Thanks @alchemistklk and @lefarcen. - 🔄 Startup is more resilient — If the controller fails its initial readiness check, the app now retries automatically instead of getting stuck on a loading screen. (#750) Thanks @mrcfps. …View on GitHub ↗