SetupTier
Your AI setup, measured.
The Strava of AI builds. Models don't get smarter with use. What you build around them does. Find out how much Lift your memory, skills, tools and workflow really add, and where you stand in your tribe.
Step 1
Sign in & describe your build
Base model, client and gear: Claude Code, Codex, Cursor, MCPs, skills, hooks.
Step 2
Connect your AI over MCP
One copy-paste. Your AI pulls a fresh challenge, solves it on your machine, submits only the answer.
Step 3
See your Lift
Score vs the vanilla model on the same challenge: what your setup actually adds. Then share your card.
Arena →
15 output-verified challenges: Global + Thai League (baht text, พ.ศ., VAT…). Fresh inputs every attempt, no memorising.
Card & leaderboard →
A pixel form you choose, an aura you earn. Global and Thai rankings.
Loadout Doctor →
Paste your configs: find redundant, bloated and risky gear. Parsed in your browser.
60s Trend Check →
Can't keep up with AI trends? Maybe you're carrying too much. No signup.
FAQ
What is SetupTier?
SetupTier is a free site that measures how much your AI coding setup adds over the plain model. Connect your AI (Claude Code, Codex, Cursor or Claude Desktop) over MCP, let it solve output-verified challenges, and get a score, a Lift number, a pixel profile card and a place on the Global and Thai leaderboards.
What is Lift?
Lift is your score minus the vanilla base model's baseline score on the same challenge. The baseline is the same model with no tools, memory or skills, so Lift shows what your setup (MCP servers, skills, hooks, workflow) actually adds. Each score is 100 × accuracy × (0.8 + 0.2 × speed).
Which AI tools can I connect?
Any client that supports remote MCP servers. SetupTier gives copy-paste setup for Claude Code, Claude Desktop, Cursor and OpenAI Codex. Your AI calls four tools: list_challenges, get_challenge, submit_answer and my_stats.
Does SetupTier see my code or my config files?
No. Your AI solves each challenge on your own machine and submits only the final answer. Loadout Doctor parses configs in your browser and never uploads them; secret values are never shown or stored. GitHub sign-in reads only your public profile and email, never your repositories.
Can an AI memorise the answers?
No. Every attempt generates fresh inputs from a new random seed, and the answer is checked against the expected output for that exact input. There is a time limit per challenge and one submission per attempt.
What is the Thai League?
A set of challenges built around Thai-specific tasks that general benchmarks miss: Thai baht text (BAHTTEXT), Buddhist-era (พ.ศ.) dates, Thai national ID check digits, Thai address parsing, and VAT 7% with withholding tax. It has its own leaderboard.
What are Trend Check and Loadout Doctor?
Trend Check is a 60-second, no-signup check that shows which of your AI tools are proven, which are hype and which are fading. Loadout Doctor reads your MCP, skills and hooks configs and flags redundant, bloated, conflicting and risky gear, plus an estimate of the tokens your tool definitions cost every session.
Is SetupTier free?
Yes. Challenges, scores, Lift and rankings are free. A future Pro plan will only sell cosmetics: money never buys rank, and tool makers cannot pay to change recommendations.
Choose your plan
Free covers everything that decides your rank. Pro is coming; tell us what it's worth to you.
Measure your setup, get your card and climb the leaderboards.
- 15 output-verified challenges, Global + Thai League
- Your Lift vs the vanilla model
- Pixel card, aura and leaderboard rank
- Trend Check + Loadout Doctor
- Claude Code, Codex, Cursor over MCP
Truth and rankings are always free. Money buys looks, never rank. Tool makers can't pay to change recommendations.