Documentation
Connecting AI coding assistants to Neuron via MCP.
Neuron ships an MCP (Model Context Protocol) server that exposes 120+ tools for managing every aspect of the platform from AI coding assistants like Claude Code, Cursor, Windsurf, and other MCP-compatible clients. Instead of using the dashboard UI, you can manage bots, conversations, knowledge bases, contacts, broadcasts, campaigns, and more — all from your IDE or terminal.
Neuron hosts the MCP server at https://mcp.neuron.ng/mcp using the Streamable HTTP transport. No installation is needed — just point your MCP client to this URL. After connecting, use the neuron_login tool to authenticate securely via your browser — no credentials are shared with the AI.
To use Neuron tools in Claude Code, add the MCP server to your project or global configuration. Create or edit the .claude/mcp.json file:
After connecting, call the neuron_login tool with no arguments. It will open a browser URL where you approve access securely — your credentials are never shared with the AI.
.claude/mcp.json
{
"mcpServers": {
"neuron": {
"type": "url",
"url": "https://mcp.neuron.ng/mcp"
}
}
}In Cursor, go to Settings > MCP Servers and add a new server. Select "Streamable HTTP" as the type and set the URL to https://mcp.neuron.ng/mcp. For Windsurf, open Settings > AI > MCP Servers and add the same URL configuration. Both IDEs support HTTP MCP servers.
The MCP server organizes tools into categories covering the entire platform: Authentication (secure browser login, MCP tokens, token refresh, org switching, logout), Bots (CRUD, pause/resume, analytics), Bot API Keys (create, list, revoke), Bot API (chat, send, conversations), Conversations (list, messages, takeover, release, close), Knowledge Bases (CRUD, entries, documents, search, sync), Tools (CRUD, secrets, testing), Webhooks (inbound CRUD, logs), Outbound Webhooks (CRUD, logs), Reflections (list, approve, reject, sync), Channels (CRUD, sessions, QR, assign bots), Contacts (CRUD, import/export, semantic search, sync), Broadcasts (CRUD, send, recipients, tracking), Campaigns (CRUD, marketplace, join, fund, send), Wallets (balance, transactions, funding), Payouts (bank account, payout requests), Pool / Marketplace (browse, publish, install), Organizations (settings, members, invitations), Billing (plans, checkout), Media (upload files), and AI (rewrite text).
The neuron_login tool supports three authentication methods. Browser auth (recommended): Call neuron_login() with no arguments — a URL opens in your browser where you approve access. Your credentials are never exposed to the AI. The token is valid for 30 days. MCP token: Generate a long-lived token from Settings > MCP Tokens in the dashboard, then pass it with neuron_login({ token: "your-token" }). Useful for environments where browser auth is unavailable. Legacy email/password: Call neuron_login({ email: "...", password: "..." }). This works but exposes credentials to the AI — use browser auth or MCP tokens instead.
For unattended use (CI/CD, scripts), generate an MCP token from the dashboard and set it via the NEURON_AUTH_TOKEN environment variable. For interactive IDE use, browser auth is the most secure and convenient option.