Give Windsurf's Cascade agent every model and tool through one MCP server
Add the BlockRun MCP server to Windsurf's mcp_config.json and Cascade can call any frontier model, generate media and read live data — no provider keys, pay per call.
How it works
Windsurf loads MCP servers from a per-user mcp_config.json under its Codeium directory. The BlockRun entry is the same JSON Claude Desktop and Cursor use: one command, no key. Cascade then has the whole catalog as tools — chat with any model, image and video generation, neural and live web search, prediction-market and market data, chain RPC.
Payment happens per request from a wallet the server creates on first run and keeps on your machine; only the signed authorization for each call leaves it. There is no team key in the workspace to leak and no seat to buy, and a model that is down fails over to a comparable one with the response saying which served.
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. ~/.codeium/windsurf/mcp_config.json
{
"mcpServers": {
"blockrun": { "command": "npx", "args": ["-y", "@blockrun/mcp@latest"] }
}
}Questions
- How do I add BlockRun to Windsurf?
- Paste the mcpServers entry above into Windsurf's mcp_config.json, reload MCP servers from Cascade's settings, and the BlockRun tools appear. The wallet is created on the first call.
- Which models can Windsurf's agent reach through BlockRun MCP?
- Every chat model on the catalog — OpenAI, Anthropic, Google, DeepSeek, xAI and the rest — plus image, video and speech generation and the data tools, all through the one server.
- Does BlockRun replace Windsurf's own model access?
- No. It adds a tool set Cascade can call; Windsurf's built-in models keep working. Use BlockRun when the agent needs a model, a modality or a data source the editor does not ship.
- Is my code or prompt stored by BlockRun?
- No. A prompt is forwarded to the provider you selected to generate the response and is not kept beyond the request or shared.
- BlockRun MCP in Windsurf vs Cursor or Cline?
- The same server and the same JSON entry in every one of them; only the config file's location differs. Pick the editor, not the integration.
- What does a call cost through this integration?
- The provider's published token rate with no markup on tokens, plus a flat per-call fee, quoted in dollars in the 402 before the call runs. Image, video and speech calls are priced per unit the same way. The live rate card is on the models page.
- Do I need an API key or an account?
- No. On the pay-per-call door the request carries its own payment from a wallet the SDK or MCP server creates on first run; there is nothing to sign up for. Teams that prefer keys get an API key and a monthly invoice instead.