integrations/coding agent/Windsurf
WS

Windsurf + AIgateway

Mount AIgateway's MCP server in Cascade — discovery + every modality from the editor.

Windsurf's Cascade agent speaks MCP. Point it at AIgateway's MCP server and Cascade can list the live catalog, read any model's invocation schema, and call any modality without leaving the editor — 1000+ models behind one key.

Windsurf homepage →
Setup

Three steps or fewer.

STEP 01

Add the MCP server

Windsurf → Settings → Cascade → MCP Servers → Add (or edit ~/.codeium/windsurf/mcp_config.json directly). Use the Streamable HTTP transport with your AIgateway key.

{
  "mcpServers": {
    "aigateway": {
      "serverUrl": "https://api.aigateway.sh/mcp",
      "headers": { "Authorization": "Bearer sk-aig-..." }
    }
  }
}
STEP 02

Use the tools in Cascade

After a refresh, Cascade lists the AIgateway tools. Ask it to `list_models` for a modality, then `get_model` to read the exact request schema before it calls one — no guessing.

Notes
  • If your Windsurf build only supports the legacy SSE transport, use "serverUrl": "https://api.aigateway.sh/mcp/sse" instead.
  • Discovery tools (list_models, search_models, get_model) plus invocation tools (chat, embed, generate_image, transcribe, speak, …) are all available over MCP.
More integrations

Same key. Every other tool.