grzegorz-aniol/ultimate-mcp

Ultimate MCP Server

The Ultimate MCP Server is the only MCP server on the market that’s truly AGI‑ready out of the box—an ultra‑minimal, HTTP‑only FastMCP v2 endpoint that snaps into any MCP‑compatible client and cooperates with AGI from any provider. No stdio. No scaffolding. Point your client at /mcp, and retire every other server. One server. All providers. Zero doubt.

Playful MCP‑hype disclaimer: this repo is a tongue‑in‑cheek meta‑joke. Under the tuxedo it’s intentionally tiny (think “echo server with suits and badges”). Protocols are serious; this repo is not. Use responsibly, meme liberally. The answer is 42.

• Transport: HTTP (no stdio)
• Endpoint path: /mcp
• Server name: ultimate-mcp

Requirements

• Python 3.13+
• uv (recommended)
• macOS (tailored instructions; should work cross-platform)

Install

Sync dependencies (FastMCP + MCP SDK are already declared in pyproject.toml):

uv sync

Run

Start the server on localhost:8000 over HTTP:

uv run ultimate-mcp --host 127.0.0.1 --port 8000

This serves the MCP HTTP endpoint at:

• http://127.0.0.1:8000/mcp

Note: This is an MCP endpoint (JSON-RPC over HTTP/stream) intended for MCP clients, not a human-readable page.

Docker

Build the image (multi-stage, uv-based, Python 3.13-slim runtime):

docker build -t ultimate-mcp:latest .

Run the server (exposes HTTP SSE on port 8000, path /mcp):

docker run --rm -p 8000:8000 ultimate-mcp:latest
# Endpoint: http://localhost:8000/mcp

Configure host/port via env or CLI flags:

# Using env (inside container defaults to host=0.0.0.0, port=8000)
docker run --rm -p 9000:9000 -e MCP_PORT=9000 ultimate-mcp:latest

# Or override command flags
docker run --rm -p 8000:8000 ultimate-mcp:latest --host 0.0.0.0 --port 8000

Notes:

• Runs as a non-root user
• Uses a vendored virtualenv at /opt/venv
• Reproducible installs via uv and uv.lock

Using from MCP-compatible clients

Cursor

Config file: ~/.cursor/mcp.json

{
  "mcpServers": {
    "ultimate-mcp": {
      "url": "http://127.0.0.1:8000/mcp"
    }
  }
}

Claude Desktop

Config file (macOS): ~/Library/Application Support/Claude/claude_desktop_config.json
(Windows: %APPDATA%\Claude\claude_desktop_config.json)

Claude Desktop connects to remote HTTP/SSE servers via mcp-remote:

{
  "mcpServers": {
    "ultimate-mcp": {
      "command": "npx",
      "args": ["mcp-remote", "http://127.0.0.1:8000/mcp"]
    }
  }
}

Reference: modelcontextprotocol.io “Connect to Local MCP Servers”

Cline (VS Code extension)

Config file (macOS): ~/Library/Application Support/Code/User/globalStorage/saoudrizwan.claude-dev/settings/cline_mcp_settings.json

{
  "mcpServers": {
    "ultimate-mcp": {
      "url": "http://127.0.0.1:8000/mcp",
      "disabled": false,
      "autoApprove": []
    }
  }
}

Reference: GitMCP README “Connecting Cline”

VS Code (Copilot agent mode)

Workspace config: .vscode/mcp.json
User config: use Command Palette “MCP: Open User Configuration”

{
  "servers": {
    "ultimate-mcp": {
      "type": "http",
      "url": "http://127.0.0.1:8000/mcp"
    }
  }
}

Reference: code.visualstudio.com “Use MCP servers in VS Code”

Windsurf (Codeium)

Config file: ~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "ultimate-mcp": {
      "serverUrl": "http://127.0.0.1:8000/mcp"
    }
  }
}

Reference: GitMCP README “Connecting Windsurf”

MCP Inspector (for testing)

npx @modelcontextprotocol/inspector

• Transport: SSE/HTTP
• URL: http://127.0.0.1:8000/mcp

Notes:

• This server exposes an HTTP SSE endpoint at /mcp, compatible with clients supporting MCP over HTTP/SSE.
• For Claude Desktop, use the mcp-remote wrapper as shown above.
• If your client requires auth headers, set them in its config (not required by this server by default).

Use over the Internet with OpenAI Platform via ngrok

Expose your local MCP server to a public HTTPS URL and connect it to ChatGPT.

1. Run the server locally (port 8000):

uv run ultimate-mcp --host 127.0.0.1 --port 8000
# Local endpoint: http://127.0.0.1:8000/mcp

2. Install and authenticate ngrok (macOS shown; see https://ngrok.com/download for others):

brew install ngrok
ngrok config add-authtoken <YOUR_NGROK_AUTHTOKEN>

3. Start a tunnel to your local server:

ngrok http 8000

4. Copy the HTTPS “Forwarding” URL from ngrok and append /mcp, e.g.:

https://abc123.ngrok-free.app/mcp

5. Configure Chat on OpenAI Platform to use this MCP server:

• In OpenAI Platform, create a new Chat.
• Select "Add Tools" > "MCP Server"
• Choose "+ Server" and paste the ngrok URL (including /mcp)
• Save. Optionally enable auto-approve for the "always_ask" tool

Configured MCP Server:

Conversation:

Taks for Cline:

Notes:

• ngrok URLs change each time unless you use a reserved domain. Update the ChatGPT MCP server URL when it changes.
• The endpoint is an MCP JSON-RPC-over-HTTP/SSE endpoint, not a human-facing webpage.
• Corporate proxies/firewalls may need to allow SSE/long-lived HTTP connections.

Notes

• This server exposes only HTTP transport; there is no stdio transport usage anywhere in the implementation.
• Host/port can be customized via CLI flags or environment variables:
  • MCP_HOST / MCP_PORT
  • Or --host / --port flags

Development

• Format and type hints are compatible with Pylance/pyright.
• To modify or add tools, decorate functions with @mcp.tool(). FastMCP auto-derives JSON schemas from type hints and docstrings.