BlockRun
Get started
Integrations / Cursor

Use BlockRun inside Cursor through its MCP config

Point Cursor's mcp.json at the BlockRun MCP server and the editor's agent can call any frontier model, generate media and pull live data, paid per request in USDC.

Get your agent on BlockRun Read the docs

How it works

Cursor loads MCP servers from a per-user mcp.json. The BlockRun entry is the same one Claude Desktop uses, so an agent inside the editor gets the whole catalog — every chat model, image and video generation, web search, market data and RPC — without a provider key in the workspace.

Because payment happens per request from a local wallet, there is no shared team key to leak into a repo and no monthly seat to buy. The wallet is created on first run, funded with USDC or by card, and only the signed authorization for each call leaves the machine.

Behind it is the BlockRun catalog: 78 chat models plus image, video, music and speech generation, search, market data and multi-chain RPC, priced per call. 6 open-weight models are free and need no wallet.

Setup

1. ~/.cursor/mcp.json (macOS/Linux) or %APPDATA%\Cursor\mcp.json (Windows)

{
  "mcpServers": {
    "blockrun": { "command": "npx", "args": ["-y", "@blockrun/mcp@latest"] }
  }
}

Questions

Does this replace Cursor's own model access?
No. It adds BlockRun as an MCP tool set the agent can call; Cursor's built-in models keep working as before.
What does a call cost?
The provider's published token rate plus a flat per-call fee, quoted in dollars in the 402 before the call runs. The live rate card is on the models page.

Related