# Fitllm MCP server

Source: https://glimind.com/server/mcp-registry%2Frun.fitllm%2Ffitllm
Registry id: `mcp-registry/run.fitllm/fitllm`
Status: live (checked 2026-10-27T01:49:46.846Z)

## Verdict

Fitllm MCP server answered 100% of Glimind's liveness checks over 30 days, better than 42% of MCP servers, with 1034 ms median handshake and 1 outage in 90 days. Its agent-readiness grade is A (95/100) across 3 tools, costing an agent about 996 tokens to load. Checked every few minutes; no tool is ever called.

## Measurements (30 days)

| Metric | Value | Better than |
|---|---|---|
| Uptime, 30 days | 100% | 42% of MCP servers |
| Median handshake latency | 1034 ms | 19% of MCP servers |
| p95 handshake latency | 3067 ms | — |
| Checks in 30 days | 4,972 | — |
| Outages, 90 days | 1 | — |
| Median recovery | 4 min | — |
| Agent-readiness score | 95/100 (A) | 80% of MCP servers |
| Tools | 3 | — |
| Tool-list size for an agent | ~996 tokens | 73% of MCP servers |
| Days observed | 29 | — |

## Install

```json
{
  "mcpServers": {
    "fitllm-mcp-server": {
      "type": "http",
      "url": "https://fitllm.run/api/mcp"
    }
  }
}
```

Claude Code:

```sh
claude mcp add --transport http fitllm-mcp-server https://fitllm.run/api/mcp
```

## Tools (3)

- `check_llm_fit`: Check whether a specific local LLM fits in the memory of a specific GPU or Apple Silicon Mac. Returns fits/tight/won't-fit verdict with the memory breakdown (we
- `list_supported`: List the built-in model names and hardware names this fit-checker knows (for mapping user wording to exact names). Standard text-only HuggingFace transformer co
- `what_fits_on_hardware`: Rank which popular local LLMs fit on a given GPU or Apple Silicon Mac (at ~4-bit quantization, 8K context) — models that fit come first, biggest first, with max

Measured by Glimind (https://glimind.com): MCP handshake + tool list only, no tool is ever called. CC BY 4.0.