Create alongside your agents
More than a chat window. Nerve Center is a connected workspace built on your OpenClaw gateway — where your ideas flow between thought, research, chat, files, and tasks. No app-switching. No copy-paste between tabs. Just you and your agents, building together.
📸 Screenshots from v0.3.5 — showing the current layout with Thoughts v2, Create mode, Research view, and Kanban.
Click any image to view full size
Quick Start • Features • Panels • First Launch • Why Nerve Center • Architecture • Credits
curl -fsSL https://raw.githubusercontent.com/PfhorScore/Nerve-Center/main/install.sh | bashRuns on: Node.js 22+ and a running OpenClaw gateway.
git clone https://github.com/PfhorScore/Nerve-Center.git
cd Nerve-Center
npm install
npm run build
npm run setup# Check version
node server-dist/index.js --version
# Start Nerve Center
node server-dist/index.js
# Open in your browser
open http://localhost:3080Nerve Center is built around a simple idea: your ideas shouldn't be trapped in one mode. Jot thoughts while your agent researches in the background. Send them to chat when you're ready. Refine into a file, a task, or the next research query — all without leaving the page. The modes connect. The work flows.
Flow between modes with your agents in tow
- Work — Chat with agents, create and edit files, browse the workspace in real time
- Research — Deep research mode generates reports with inline citations, sources, and images
- Create - Work with Markdown,.txt, and website designer with live preview
- Tasks — Full Kanban board with AI-assisted card management and agent-driven workflows
Your main workspace — chat, create, and edit alongside your agents.
- Chat — Talk to your agents with streaming responses, tool calls, and rich markdown
- Split view — Chat alongside your documents. Open any file and keep the conversation going beside it. Drag the divider to resize, or collapse the chat for full document focus
- Panels — Workspace (file tree), Thoughts (notes), Library (references), Activity (agent logs)
Deep research interface for your agent
- Quick & Deep search modes using any OpenClaw-compatible model provider
- Rich markdown answers with inline citation links
[1], sources, and images - AI auto-sort — one click splits a conversation into topic-based threads
- Hover previews on every source — favicon, title, snippet at a glance
- Tabbed results: All, Sources, Images, Links
- Follow-up suggestions — click to dive deeper without retyping
- Thread sidebar with agent-generated titles that evolve with your discussion
A full Kanban board for managing work alongside your agent. Not a toy — real columns, real workflows.
- Drag-and-drop cards between columns (Backlog, In Progress, Review, Done)
- AI-assisted card creation — describe the task, the agent helps flesh it out
- Task detail drawer — full description, linked files, agent assignments, status history
- Proposal inbox — agents can propose tasks for your approval
- Quick view — peek at task details without leaving the board
- Configurable columns — adapt the workflow to match your process
Keep tabs on your agents without context-switching.
- Agent Hub — Central drawer for moderating agents, browsing memory, managing settings
- Analytics Dashboard — Token usage charts, memory list, usage limits at a glance, accessible from the Agent Hub
- Full Memory Browser — add, edit, delete, and search agent memories. Not just read-only — you can refine what your agent remembers
- Charts & data — inline charts, lightweight financial charts, and TradingView widget support for data-rich conversations
- Agent system monitoring — connection status, gateway health, session list
One of the few open-source AI tools with full real-time voice support — speak to your agents and hear them respond, all hands-free.
- Speech-to-text — Dictate messages using local Whisper (offline, private) or OpenAI Whisper
- Text-to-speech — Agent replies read aloud via Edge TTS (free), OpenAI TTS, or Replicate
- Push-to-talk — Double-tap Shift to start recording, press again to send. No buttons needed
- Live transcription — See your speech transcribed in real time before it sends
- Voice toggle — Disable voice replies mid-response — cuts audio immediately
Extend agents with third-party tools via the Model Context Protocol.
- Server Manager — Register, enable, and disable MCP servers from Settings
- Tool call visibility — See and toggle MCP tool blocks in chat
- Inline embedded apps — Interactive MCP content renders inline in the chat stream
- Doctor CLI — Run
npm run doctorto check gateway connectivity, server health, and model readiness - One-Click Updates — Update Nerve Center from the UI with real-time progress streaming
- Hover Commands — Hover over messages for copy, think, research, and read-aloud actions
- Command Palette —
Cmd+Kto quickly switch panels, views, and create files - Notifications — Subtle audio cues for message completion and errors
From Thoughts 🧠, to brainstorming 💡, to research 🔬, and back again.
- Think about this? — Hover any message and send it to the Thoughts panel to capture a note
- Research this — Highlight text or hover a message to kick off deep research
- Discuss this — Send a thought straight into the conversation
- Flow between modes without context-switching. Your agent follows you between Work, Research, and Tasks.
Nerve Center's panels aren't a static dashboard — they're stations in a workflow. Capture in Thoughts. Discuss in Chat. Research deeper. Save to files. Track with tasks. Each one feeds the next, and your agent moves alongside you. Drag any panel between sides, resize, or collapse to icon strips — set up your flow how it makes sense for what you're doing right now.
Edit files and browse your agent's workspace without leaving the chat. Full file management in the sidebar:
- File tree browser with folder navigation and context menus
- Built-in file editor with syntax highlighting (CodeMirror)
- Code editing for markdown, JavaScript, Python, and more
- Smart workspace root — auto-detects agent workspace paths
- Hidden file support — toggle visibility for dotfiles
- Create, rename, move, and delete files directly from the panel
- File watcher — live updates when files change on disk
Ctrl+Entersplits your notes into individual cards- Check off completed thoughts (dimmed, stays for reference)
- Auto-detect completion — send to chat, auto-checks when the AI finishes
- Hover actions — copy, send to chat, or research each thought
- Click to edit any thought inline
- Server-backed sync via
Thoughts.md— notes available across all devices - Multi-select — toggle select mode, batch select thoughts with Shift+click, send multiple thoughts to chat at once
- Attachments — attach images and files to thoughts for richer notes
- Completion state syncs to server — the agent (or other devices) can check off your thoughts
Every link, citation, and image, auto-organized. All your chat references in one place:
- Auto-extracts all URLs, citations, and images from messages
- Deduplicated by URL — clean, no clutter
- Tabs for All / Links / Images with live counts
- Search to filter specific references
- Favicon previews and image thumbnails
Watch your agents work without the chat clutter. Tool calls grouped by message (collapsed by default):
- Shows tool name, description, arguments, and status (running ✓ error)
- Jump to message — click the icon to scroll to the corresponding chat entry
- Finished activity persists in an archive for later review
- Separate from the chat stream — clean conversations
Your fleet at a glance. Switch, browse, configure. Central drawer for managing your agent fleet:
- Agent selector — switch between agents, view session status
- Memory browser — read and search agent memories
- Settings — configure TTS, STT, theme, panel layout, and more
- Avatars — upload per-agent profile images for chat headers and session list
- System monitoring — token usage, connection status, gateway health
Tip: Customize the layout further with collapsible sidebars, drag-and-drop panel reordering, and resizable panels.
Once Nerve Center is running (node server-dist/index.js or via the install script), here's your first 5 minutes:
- Open your browser to
http://localhost:3080 - Nerve Center auto-detects your gateway — if OpenClaw is running, you're authenticated
- Pick an agent from the Agent Hub (icon in the top-right toolbar)
- Say hello — type a message and hit send. Watch the response stream in real time
- Try research — click the Research tab in the view switcher and ask a deep question
- Open the panels — Workspace on the left, Library on the right. Drag them between sides to see how panels move
- Hit
Cmd+K(orCtrl+K) — the command palette pops up. Search for "thoughts" to jump to the Thoughts panel - Save your thoughts — type something in the Thoughts panel and press
Ctrl+Enterto split it into a card
Pro tip: If you're accessing from another machine (like your phone), set
HOST=0.0.0.0in.envor use Tailscale for a secure tunnel.
| Symptom | Fix |
|---|---|
| Browser says "Connection refused" | Make sure OpenClaw gateway is running (openclaw gateway status) |
| Port 3080 already in use | Change PORT in .env and restart |
NERVE_AUTH locked me out |
Check the password you set in .env |
| Can't reach from another device | Set HOST=0.0.0.0 in .env or connect via Tailscale |
| Research tab shows no models | Make sure your gateway has at least one model provider configured |
| Panels look wrong or missing | Try a hard refresh (Ctrl+Shift+R). Layout resets to defaults if you clear localStorage |
Nerve Center is a feature fork of the original Nerve. It adds capabilities you won't find in the upstream:
| Feature | Nerve Center |
|---|---|
| Deep Research Tab | ✅ Full Perplexity-class research with AI auto-sort, citations, threads |
| Kanban & Tasks | ✅ Drag-and-drop board, AI-assisted cards, proposal inbox |
| MCP Integration | ✅ Server manager, tool call visibility, inline embedded apps |
| Dashboard & Monitoring | ✅ Token usage charts, memory browser, agent health |
| Charts & Data Viz | ✅ Inline charts, lightweight charts, TradingView widgets |
| Workspace Panel | ✅ File tree, code editor, live file watcher |
| Thoughts Panel v2 | ✅ Card-based notes with completion tracking, server sync |
| Library Panel | ✅ Auto-extracted references, URLs, images from chat |
| Activity Panel | ✅ Live agent activity separate from chat stream |
| Agent Hub | ✅ Central drawer for agents, memory, settings, avatars |
| Clean Input Bar | ✅ Buttons below text, file upload, image lightbox |
| One-Click Updates | ✅ Update & restart from the UI with progress |
| Auth & Security | ✅ Password login, gateway token auth |
| Node Pairing | ✅ Connect remote machines via Tailscale or LAN |
| Collapsible sidebars | ✅ VS Code-style hover-to-expand |
| Drag-and-drop layout | ✅ Move and reorder panels between sides |
| Avatars | ✅ Per-agent profile images |
| Discord-style messages | ✅ Clean username + avatar + timestamp layout |
| Research view | ✅ Dedicated fullscreen research workspace |
| All of this sits on top of the rock-solid OpenClaw gateway and agent infrastructure. |
Browser ─── Nerve Center (:3080) ─── OpenClaw Gateway (:18789)
│ │
├─ WS ───────┤ proxied to gateway
├─ SSE ──────┤ file watchers, real-time sync
└─ REST ─────┘ files, memories, TTS, models
Frontend: React 19 · Tailwind CSS 4 · shadcn/ui · Vite 7 Backend: Hono 4 on Node.js 22+
Configure Nerve Center through a .env file in the project root. Key settings:
| Variable | Default | Description |
|---|---|---|
PORT |
3080 |
HTTP listen port |
HOST |
127.0.0.1 |
Bind address (0.0.0.0 for LAN/remote) |
NERVE_AUTH |
false |
Enable password-protected login |
GATEWAY_TOKEN |
— | Your OpenClaw gateway auth token |
AGENT_NAME |
Agent |
Display name for the agent |
Run npm run setup for an interactive configuration walkthrough. |
git clone https://github.com/PfhorScore/Nerve-Center.git
cd Nerve-Center
npm install
# Start the dev server with hot reload
npm run dev
# Or build for production
npm run build
node server-dist/index.js