MCP Directory

How to add Roundtable AI MCP Server to Windsurf

Local MCP server that delegates coding tasks to specialized AI sub-agents (Codex, Claude, Cursor, Gemini) in parallel. Paste the config into ~/.codeium/windsurf/mcp_config.json and restart Windsurf.

Last updated June 14, 2026 · 115★ · stdio · no auth

Windsurf config for Roundtable AI MCP Server

pip install roundtable-ai
{
  "mcpServers": {
    "roundtable-ai-mcp-server": {
      "command": "uvx",
      "args": [
        "roundtable-ai@latest"
      ],
      "env": {
        "CLI_MCP_SUBAGENTS": "codex,claude,cursor,gemini"
      }
    }
  }
}

Requires `uv` (the Python package runner). Install it from https://docs.astral.sh/uv/ if `uvx` is not found.

Setup steps

  1. 1Open Windsurf → Cascade → the hammer/MCP icon → Configure (or edit ~/.codeium/windsurf/mcp_config.json).
  2. 2Paste the Roundtable AI MCP Server config below.
  3. 3Fill in placeholder secrets, then save.
  4. 4Click Refresh in the MCP panel.
  5. 5Roundtable AI MCP Server's tools become available to Cascade.

Before you start

  • Python 3.10+
  • pip or uv/uvx to install the roundtable-ai package
  • The CLI tools for the sub-agents you enable (e.g. Codex, Claude Code, Cursor, Gemini), installed and authenticated with your own API keys/subscriptions

What Roundtable AI MCP Server can do in Windsurf

check_codex_availability

Verify Codex CLI status / availability.

check_claude_availability

Verify Claude Code CLI status / availability.

check_cursor_availability

Verify Cursor CLI status / availability.

check_gemini_availability

Verify Gemini CLI status / availability.

execute_codex_task

Run coding tasks through Codex.

execute_claude_task

Run coding tasks through Claude Code.

execute_cursor_task

Run coding tasks through Cursor.

execute_gemini_task

Run coding tasks through Gemini.

Security

Runs locally and invokes your installed CLI coding agents (Codex, Claude Code, Cursor, Gemini), executing tasks against your codebase with your existing API keys/subscriptions. Sub-agents run local tools such as read_file and run_shell_command, so only enable agents you trust with shell and file access in your project.

Roundtable AI MCP Server + Windsurf FAQ

Where is the Windsurf config file?

Windsurf reads MCP servers from ~/.codeium/windsurf/mcp_config.json. Paste the Roundtable AI MCP Server config there under the "mcpServers" key and restart the client.

Is Roundtable AI MCP Server safe to use with Windsurf?

Runs locally and invokes your installed CLI coding agents (Codex, Claude Code, Cursor, Gemini), executing tasks against your codebase with your existing API keys/subscriptions. Sub-agents run local tools such as read_file and run_shell_command, so only enable agents you trust with shell and file access in your project.

Does Roundtable charge extra to use multiple models?

No. It invokes the CLI tools you already have installed and uses your existing API keys and subscriptions with zero markup; cost is the same as running the tools manually.

How do I choose which AI assistants are enabled?

Set the CLI_MCP_SUBAGENTS environment variable (e.g. "codex,claude,cursor,gemini") or pass --agents on the command line (e.g. `roundtable-ai --agents codex,claude`).

How do I add it to Claude Code?

Run `claude mcp add roundtable-ai -- roundtable-ai --agents gemini,claude,codex,cursor`, or with uvx: `claude mcp add roundtable-ai -- uvx roundtable-ai@latest --agents gemini,claude,codex,cursor`.

View repo Full Roundtable AI MCP Server page