Skip to main content
Arcus exposes its documentation as a remote Model Context Protocol (MCP) server. Once connected, your AI assistant can search the REST and WebSocket reference, surface examples, and answer questions like “where do I set TP/SL?” without leaving your editor.

Server details

No sign-in, API key, or token is needed: add the URL to your MCP client and it connects. Opening the URL in a browser returns 405 Method Not Allowed, because MCP clients talk to it with POST requests.

What you can ask

After connecting, your assistant can answer questions grounded in the live spec:

API behavior

“Which endpoint takes TP/SL — placeOrder or batchPlaceOrders?”

Auth flow

“Show me how to sign a request with my Ed25519 API key.”

WebSocket channels

“What channels stream order book updates and what do their payloads look like?”

Code samples

“Give me a TypeScript snippet that cancels all open orders for one market.”

Connect your editor

Claude Code

Remote MCP via claude mcp add.

Cursor

Add through Cursor MCP settings.

Windsurf

Configure in Cascade’s MCP panel.

Troubleshooting

Expected. The endpoint only answers MCP requests (POST), so a browser GET is refused. Use a real MCP client (Claude Code, Cursor, Windsurf, Claude Desktop).
Make sure your client supports remote MCP servers over streamable HTTP, and update it to the latest version if not. For Claude Code: npm i -g @anthropic-ai/claude-code. Then remove and re-add the server.
The MCP server reflects the current published spec. If something looks wrong, open an issue and link to the page — we resync from upstream on every merge.