Cline & Other MCP Clients

In addition to Claude Desktop and Cursor, cotbias-mcp is compatible with any client implementing Anthropic's Model Context Protocol (MCP) over standard input/output (stdio).

---

🛠 Cline (VS Code Extension) Setup

1. Install the Cline Extension in VS Code.
2. In the Cline settings tab, click the MCP Servers icon (or open cline_mcp_settings.json).
3. Add the cotbias server configuration:

{
  "mcpServers": {
    "cotbias": {
      "command": "uvx",
      "args": [
        "cotbias[mcp]",
        "cotbias-mcp"
      ],
      "env": {
        "COT_API_KEY": "cot_live_YOUR_ACTUAL_API_KEY"
      }
    }
  }
}

1. Save the file. Cline will automatically start the server and register the tools.

---

🌊 Windsurf IDE Setup

In Windsurf, open your user configuration (~/.codeium/windsurf/mcp_config.json) and add the same configuration block under mcpServers.

---

💬 LibreChat Setup

In your librechat.yaml configuration file:

mcpServers:
  cotbias:
    type: stdio
    command: uvx
    args:
      - "cotbias[mcp]"
      - "cotbias-mcp"
    env:
      COT_API_KEY: "cot_live_YOUR_ACTUAL_API_KEY"

---

⚠️ Note on Hosted Web Clients (ChatGPT, Gemini)

Local Stdio vs Hosted Remote Server

Web-based LLM interfaces such as ChatGPT (OpenAI) and Gemini Web do not currently support local stdio subprocess execution on your desktop machine.

They require a public, hosted HTTPS remote server with OAuth2 authentication. A hosted remote MCP endpoint for web-based AI clients is on the product roadmap for a future release. For now, use cotbias-mcp with desktop agents (Claude Desktop, Cursor, Cline, Windsurf).