llmcloud.ai
Features · MCP

Every model, as a tool call.

mcp.llmcloud.ai speaks the Model Context Protocol. Point any MCP-aware client at it and every model behind the gateway becomes a callable tool — routed by the same policies as the HTTP API.

Claude Code

Adds every llmcloud model as an ask_* tool. Great for delegating cheap subtasks off Sonnet.

claude mcp add llmcloud https://mcp.llmcloud.ai \
  --header "Authorization: Bearer $LLMCLOUD_API_KEY"

Cursor

Drop this into ~/.cursor/mcp.json — the router shows up as a tool source in the composer.

{
  "mcpServers": {
    "llmcloud": {
      "url": "https://mcp.llmcloud.ai",
      "headers": { "Authorization": "Bearer YOUR_KEY" }
    }
  }
}

Windsurf

Cascade picks up the tools on next reload. Combine with auto:cost to keep long agent runs affordable.

{
  "mcpServers": {
    "llmcloud": {
      "serverUrl": "https://mcp.llmcloud.ai",
      "headers": { "Authorization": "Bearer YOUR_KEY" }
    }
  }
}

Tools exposed

  • ask_autoDelegate a subtask under the auto:quality policy.
  • ask_cheapFastest / cheapest capable model for the input modality.
  • ask_reasoningRoute to a thinking model, return reasoning[] + final.
  • list_modelsLive catalog, filterable by modality, price, and provider.
  • web_searchGrounded search, citations returned in a standard shape.