FitLLM is a remote MCP server that gives an AI agent 3 tools, among them check_llm_fit, list_supported, what_fits_on_hardware. It answers at https://fitllm.run/api/mcp over streamable HTTP, and it listed its tools to an anonymous client, so no sign-in was needed for that. Our probes reached it 6 of 6 times since 12 September 2026; the last check, on 24 October 2026, succeeded.
Will this LLM fit on your GPU, multi-GPU rig or Mac? Exact VRAM & KV-cache math. Read-only.
Description as the publisher wrote it.
| Tool | What it does |
|---|---|
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 (weights, KV cache, linear-attention state when present, runtime overhead, reserve) Inputs: model, gpu, gpu_count, mac_ram_gb, quant, context_tokens |
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 configs can also be checked via fitllm.run; unsupported architectures are rejected |
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 context each. Use when a user asks "what can I run on my <GPU/Mac/N GB>?", "bes Inputs: gpu, gpu_count, mac_ram_gb |
Every block below points at https://fitllm.run/api/mcp. Pick your client and copy.
claude mcp add --transport http fitllm https://fitllm.run/api/mcp
{
"mcpServers": {
"fitllm": {
"command": "npx",
"args": ["-y", "mcp-remote", "https://fitllm.run/api/mcp"]
}
}
}{
"mcpServers": {
"fitllm": {
"url": "https://fitllm.run/api/mcp"
}
}
}{
"servers": {
"fitllm": {
"type": "http",
"url": "https://fitllm.run/api/mcp"
}
}
}{
"context_servers": {
"fitllm": {
"url": "https://fitllm.run/api/mcp"
}
}
}{
"mcpServers": {
"fitllm": {
"serverUrl": "https://fitllm.run/api/mcp"
}
}
}{
"mcpServers": {
"fitllm": {
"url": "https://fitllm.run/api/mcp"
}
}
}name: FitLLM
version: 0.0.1
schema: v1
mcpServers:
- name: FitLLM
type: streamable-http
url: https://fitllm.run/api/mcp{
"mcpServers": {
"fitllm": {
"httpUrl": "https://fitllm.run/api/mcp"
}
}
}{
"mcpServers": {
"fitllm": {
"type": "http",
"url": "https://fitllm.run/api/mcp"
}
}
}{
"mcpServers": {
"fitllm": {
"url": "https://fitllm.run/api/mcp"
}
}
}goose://extension?url=https%3A%2F%2Ffitllm.run%2Fapi%2Fmcp&type=streamable_http&id=fitllm&name=FitLLM
It answered our last check on 24 October 2026 in 265 ms. Over 6 checks since 12 September 2026 it answered 6. Status here comes from Neuronto's own probes, not from the publisher.
Yes. It is published by fitllm.run and served from fitllm.run, the vendor's own domain: https://fitllm.run/api/mcp.
3, as returned by the server's own tools/list on 23 September 2026: check_llm_fit, list_supported, what_fits_on_hardware.
It answered our last check on 24 October 2026 in 265 ms. Over 6 checks since 12 September 2026 it answered 6.
Put the server under mcpServers in ~/.cursor/mcp.json with the URL https://fitllm.run/api/mcp. The exact JSON is above, with the same setup for Claude Code, Claude Desktop, VS Code, Windsurf, Zed, Gemini CLI and the rest.
It listed its tools to an anonymous client, so reading the tool list needs no key. Individual tools can still ask for credentials in their own inputs.
Publisher: run.fitllm. Endpoint host: fitllm.run. ARD identifier: urn:air:run.fitllm:mcp:fitllm.
First indexed 10 September 2026. The tools are exactly what the server returned; nothing on this page
is inferred from its name. Ranking in Neuronto search is never for sale.