model-radar

MCP server that pings 130+ free coding LLM models across 17 providers in real-time, ranks them by latency, and helps AI agents pick the fastest available model.

Inspired by free-coding-models.

Install

pip install model-radar

Quick Start

1. Configure an API key

# Option A: Save to ~/.model-radar/config.json
model-radar configure nvidia nvapi-xxx

# Option B: Environment variable
export NVIDIA_API_KEY=nvapi-xxx

Or copy the template: cp config.example.json ~/.model-radar/config.json and edit it.

2. Add to your MCP client

Claude Code (~/.claude/settings.json):

{
  "mcpServers": {
    "model-radar": {
      "command": "model-radar",
      "args": ["serve"]
    }
  }
}

Cursor (.cursor/mcp.json in project root or ~/.cursor/mcp.json):

Stdio (default — Cursor starts the server):

{
  "mcpServers": {
    "model-radar": {
      "command": "/path/to/your/.venv/bin/model-radar",
      "args": ["serve"]
    }
  }
}

SSE (you run the server; Cursor connects by URL):

The server listens on one port and serves both Streamable HTTP (/mcp) and SSE (/sse, /messages/). Cursor tries Streamable HTTP first, then SSE, so it can connect as soon as the server is up.

# Terminal: start the server (leave it running)
model-radar serve --transport sse --port 8765

Then in Cursor MCP config use the URL http://127.0.0.1:8765 (or http://127.0.0.1:8765/mcp / http://127.0.0.1:8765/sse as your client expects). Start the server before opening the project so Cursor finds it immediately.

Web dashboard: With --web, the same server serves a localhost UI at http://127.0.0.1:8765/ for status, config, discovery, and running prompts (REST API at /api/*). MCP remains at /sse. Privacy: The server binds to 127.0.0.1 only; your API keys and data never leave your machine. Keys are stored only in ~/.model-radar/config.json (0o600).

model-radar serve --transport sse --port 8765 --web

Restarting the SSE server: After updating model-radar, restart the server so new tools appear. You can either restart the process manually, or run with a restart wrapper and use the restart_server() MCP tool:

# Allow the MCP tool to request exit; a loop restarts the server
export MODEL_RADAR_ALLOW_RESTART=1
while true; do model-radar serve --transport sse --port 8765; sleep 1; done

Then call the restart_server() tool (e.g. from an agent); the process exits, the loop starts a new one with updated code, and you reconnect.

OpenClaw (~/.openclaw/openclaw.json):

{
  "mcpServers": {
    "model-radar": {
      "command": "model-radar",
      "args": ["serve"]
    }
  }
}

3. CLI usage

# Scan models
model-radar scan --min-tier S --limit 10

# List providers
model-radar providers

# Save a key
model-radar configure nvidia nvapi-xxx

Providers (17)

MCP Tools

• list_providers() — See all 17 providers with config status

• list_models(tier?, provider?, min_tier?) — Browse the model catalog

• scan(tier?, provider?, min_tier?, configured_only?, limit?) — Ping models in parallel, ranked by latency

• get_fastest(min_tier?, provider?, count?) — Quick: best N models right now

• provider_status() — Per-provider health check

• configure_key(provider, api_key) — Save an API key

Tier Scale (SWE-bench Verified)

License

MIT