Benchmark local LLM models — speed, quality & hardware fitness verdict from any MCP client
Benchmark local LLM models — speed, quality & hardware fitness verdict from any MCP client
By default, production builds upload shared results to the official MetriLLM leaderboard (https://metrillm.dev).
Compare MCP servers by category, or browse hosted MCP servers if you prefer an endpoint over a local process.
{
"mcpServers": {
"metrillm": {
"command": "npx",
"args": ["-y", "metrillm"],
"env": {
"METRILLM_SUPABASE_URL": "your-value",
"METRILLM_SUPABASE_ANON_KEY": "your-value",
"METRILLM_PUBLIC_RESULT_BASE_URL": "your-value",
"OLLAMA_HOST": "your-value",
"LM_STUDIO_BASE_URL": "your-value",
"LM_STUDIO_API_KEY": "your-value"
}
}
}
}Add to claude_desktop_config.json, then restart Claude Desktop.
| Variable | Description | Required |
|---|---|---|
| METRILLM_SUPABASE_URL | Endpoint or connection string the server talks to. | Yes |
| METRILLM_SUPABASE_ANON_KEY | Credential the server authenticates with. | Yes |
| METRILLM_PUBLIC_RESULT_BASE_URL | Endpoint or connection string the server talks to. | Yes |
| OLLAMA_HOST | Endpoint or connection string the server talks to. | Optional |
| LM_STUDIO_BASE_URL | Endpoint or connection string the server talks to. | Yes |
| LM_STUDIO_API_KEY | Credential the server authenticates with. | Yes |
Enables language models to access and query athlete activity data from the Strava API.
Enables interaction with the Strava API through a Model Context Protocol (MCP) server.
A Model Context Protocol server that lets AI assistants read and analyse your Intervals.icu training data the way a coach would: activities with every custom field and stream, intervals, wellness against personal…
Exercises, muscles, and workout data from the open wger database. No key required.
Fitness Ai tools for AI agents. Capabilities: generate workout, track calories, calculate bo...
Your RIRLift strength training log: workouts, progress, volume, recovery and program drafts.