Use any model, in any MCP client.

ScoutRival is a standard MCP server, so it works with any client that speaks MCP — and the client, not ScoutRival, decides which model runs. That's how you use Gemini, DeepSeek, Kimi, a local model or any other behind the same ScoutRival tools.

The one idea worth holding on to

MCP is a feature of the client, not the AI model. The tools ScoutRival exposes are the same on every client; the client you open decides which model answers. So the question isn't "does this model support ScoutRival" — it's "does the client I'm using let me add a custom MCP server, and let me pick the model". If yes, you're set.

That flips the usual worry on its head. "Can I use ScoutRival with Gemini, DeepSeek or Kimi?" The answer is yes for all of them — you run that model inside an MCP-capable client, and ScoutRival's tools sit behind it unchanged.

Clients that fit

These clients let you add a custom or remote MCP server and choose the model that drives it. Each has its own way to add a server, so check its own docs for the exact screen:

• Cline and Continue — VS Code extensions.
• Windsurf
• Zed
• Goose
• LibreChat
• Jan
• Raycast
• VS Code — its own built-in MCP support.
• Warp

How to connect, in general

Many clients accept exactly this JSON shape:

mcpServers config{
  "mcpServers": {
    "scoutrival": {
      "url": "https://user.scoutrival.com/api/mcp",
      "headers": { "Authorization": "Bearer sr_mcp_YOUR_TOKEN" }
    }
  }
}

Others don't take pasted JSON at all — they give you separate form fields for the server URL and the header. Both routes end up in the same place: the URL above, plus the bearer token.

Model by model

• Gemini. The Gemini CLI can add a custom MCP server directly. The Gemini consumer app supports pre-built connectors only, so for a custom server like ScoutRival, run Gemini as the model inside one of the MCP clients above.
• DeepSeek and Kimi. Neither ships its own MCP connector, so use them as the backend model inside an MCP client. Many of the clients above let you set an OpenAI-compatible endpoint and API key for DeepSeek or for Kimi (Moonshot). ScoutRival's tools work the same either way.
• Grok. Grok has its own connectors on grok.com and a CLI, so it can add ScoutRival directly — no separate client needed.
• A local model. Run it — for example through a local OpenAI-compatible server — inside a client such as Cline, Goose or Jan, then add ScoutRival as the MCP server.