Real-time web cockpit for OpenClaw: voice conversations, agent automated kanban board, workspace/file control, sub-agent sessions, inline charts, and usage visibility.
# Add to your Claude Code skills
git clone https://github.com/daggerhashimoto/openclaw-nerveThe cockpit OpenClaw deserves.
OpenClaw is powerful. Nerve is the interface that makes people say “oh, now I get it."
curl -fsSL https://raw.githubusercontent.com/daggerhashimoto/openclaw-nerve/master/install.sh | bash
Run the installer, live in 60 seconds
https://github.com/user-attachments/assets/25d65a85-1d42-45bc-baae-5e6fca531705
Chat is great for talking to agents. It is not enough for operating them.
The moment you care about visibility, control and coordination over your agents, the thread gets too small. You want the workspace, sessions, taskboard, editor, usage, and agent context in one place.
Nerve is that place.
Run multiple agents from one place. Each agent can have its own workspace, subagents, memory, identity, soul, and skills, while Nerve gives you a single control plane to switch context, inspect state, and operate the whole fleet.
Push-to-talk, wake word flows, explicit language selection, local Whisper transcription, multilingual stop and cancel phrases, and multiple TTS providers. Voice is part of the product, not an afterthought.
No comments yet. Be the first to share your thoughts!
Each agent can have its own workspace, memory, identity, soul, and skills. Nerve lets you inspect, edit, and manage that context live, without guessing what an agent knows, where it works, or how it is configured.
Crons, session trees, kanban workflows, review loops, proposal inboxes, and model overrides. Nerve gives agent work an operating surface instead of leaving it trapped inside chat history.
Charts, diffs, previews, syntax-highlighted code, structured tool rendering, and streaming UI that makes agent responses easier to inspect.
- Talk to your agent by voice and hear it answer back naturally
- Browse and edit the workspace live while the conversation is still happening
- Watch cron runs as separate sessions instead of treating automation like a black box
- Delegate work onto a kanban board and review what came back
- Ask for a chart and get a real chart, not a code block pretending to be one
- Track token usage, costs, and context pressure while long tasks run
- Inspect subagent activity without losing the main thread
- Switch between per-agent workspaces and memory without losing context
- Inspect each agent’s identity, soul, and skills from the UI
- Delegate subagent work inside a larger agent fleet instead of treating everything as one thread
| Area | Highlights | |---|---| | Agent fleet | Run multiple agents from one control plane, each with its own workspace, subagents, memory, identity, soul, and skills | | Interaction | Streaming chat, markdown, syntax highlighting, diff views, image paste, file previews, voice input, TTS, live transcription preview | | Workspace | Per-agent file browser, tabbed editor, memory editing, config editing, skills browser | | Operations | Session tree, subagents, cron scheduling, kanban task board, review flow, proposal inbox, model overrides | | Observability | Token usage, cost tracking, context meter, agent logs, event logs | | Polish | Command palette, responsive UI, 14 themes, font family and 10px to 24px font size controls, mobile-safe input sizing, hot-reloadable settings, updater with rollback |
curl -fsSL https://raw.githubusercontent.com/daggerhashimoto/openclaw-nerve/master/install.sh | bash
The installer handles dependencies, clone, build, and then usually hands off straight into the setup wizard. Guided access modes include localhost, LAN, Tailscale tailnet IP, and Tailscale Serve.
git clone https://github.com/daggerhashimoto/openclaw-nerve.git
cd openclaw-nerve
npm install
npm run setup
npm run prod
npm run update -- --yes
Fetches the latest release, rebuilds, restarts, verifies health, and rolls back automatically on failure.
npm run dev # frontend — Vite on :3080 by default
PORT=3081 npm run dev:server # backend — explicit split-port dev setup
npm run dev:server uses the normal server PORT setting. If you do not override it, the backend also defaults to :3080 and will collide with Vite.
Requires: Node.js 22+ and an OpenClaw gateway.
Nerve sits in front of the gateway and gives you a richer operating surface in the browser.
Browser ─── Nerve (:3080) ─── OpenClaw Gateway (:18789)
│ │
├─ WS ──────┤ proxied to gateway
├─ SSE ─────┤ file watchers, real-time sync
└─ REST ────┘ files, memories, TTS, models
OpenClaw remains the engine. Nerve gives it a cockpit.
Frontend: React 19 · Tailwind CSS 4 · shadcn/ui · Vite 7 Backend: Hono 4 on Node.js
Nerve binds to 127.0.0.1 by default, so it stays local unless you choose to expose it.
When you bind it to the network (HOST=0.0.0.0), built-in password authentication protects the UI and its endpoints. Sessions use signed cookies, passwords are stored as hashes, WebSocket upgrades are authenticated, and trusted connections can use server-side gateway token injection.
For the full threat model and hardening details, see docs/SECURITY.md.
.env variables and setup behaviorIf this is the kind of interface you want around your OpenClaw setup, give the repo a star, contribute and keep an eye on it.
Join the Nerve Discord to get help, discuss, share your setup, and follow development.