FitMyLLM
CLI & MCP

FitMyLLM in your terminal and IDE.

Full-featured TUI for the terminal. Query the same 498-model · 1,720-GPU database from Claude Code, Cursor, VS Code via the Model Context Protocol. One API key for both.

API key

Free · 30 req/min

An API key is required to use the MCP server. Free, no limits beyond rate limiting.

Loading…

CLI tool

pip install fitmyllm

Full-featured terminal UI (Textual) for finding models, comparing GPUs, chatting. Same engine as the website.

01 · Install
$pip install fitmyllm
02 · Setup key
$export FITMYLLM_API_KEY=fml_your_key_here
03 · Run
$fitmyllm
Features · 10
Find Models
Auto-detect GPU, 11 filters, 30+ models ranked
Find GPU
GPU recommendations with budget/speed/vendor filters
Enterprise
10-tab analysis: risk, TCO, scaling, SLA, architecture
Compare
Side-by-side comparison of up to 4 models
Install
Choose quant, pick engine, install with live progress
Chat
Talk to models via Ollama with streaming + thinking blocks
Tier List
Models and GPUs ranked S-F with cloud alternatives
Benchmarks
Leaderboard sortable by 8 benchmark metrics
GPU Prices
Search and compare GPU pricing
Charts
ASCII Score/Speed/VRAM charts + scatter plot
Keyboard shortcuts
fToggle filters
gSearch/change GPU
SpaceMark for compare
cCompare / Chat
iInstall model
tCommand simulator
sSave model
rHuggingFace README
eExport as Markdown
vASCII charts
Ctrl+SSave filters
Ctrl+TToggle thinking
EscGo back
qQuit

MCP tools

5 tools
recommend_models

Find the best models for a specific GPU

"What models can I run on my RTX 4060?"

search_models

Search 498 open-source LLM models

"Search for Qwen 3.5 models"

search_gpus

Search 1700+ GPUs by name

"Find specs for RTX 5090"

get_model_details

Full model details: benchmarks, VRAM, quantizations

"Show me details for Llama 3.1 8B"

compare_models

Compare 2-4 models side by side

"Compare Qwen 3.5 27B vs Llama 3.1 70B"

How it works

MCP (Model Context Protocol) lets AI assistants call external tools during a conversation.

  1. 01You add the FitMyLLM MCP server to your IDE (one-time setup)
  2. 02When you ask something like "What model fits my RTX 4060?", the AI calls our recommend_models tool
  3. 03Our server runs the same engine as the website and returns results
  4. 04The AI uses the data to answer your question

Rate limit · 30 req/min · data updated weekly