One question. Up to four AI models. One agreement score.

Measure how much AI agrees.

Tetralyn puts one question to up to four AI models at once, compares what comes back and gives you a number for how closely the answers line up, with every point of drift named.

2 free questions daily, with no wallet and no email to start.

1
question
4
models at most, per run
0–100
agreement scale

Sample answers, for illustration

Should we rewrite this service in Rust?

78agreement

One signal drifts: team experience

Claude Opus 5

Not yet. Profile first and fix the hot paths.

GPT-6 Astra

Rewrite only the hot path, once it is measured.

Grok 4.6

Measure first. Most of it won’t need the rewrite.

DeepSeek V4 Pro

Yes, if the team already knows Rust well.

Why measure it

A fluent answer isn’t a measured one.

  1. One input, every channel

    Write the question once and each selected model receives it. No tab-hopping, no pasting.

  2. Signals land as they finish

    Each answer card fills the moment its model responds, so a slow model never holds the others back. Drag the cards into the order you want.

  3. Swap channels any time

    Run up to four models and change them between questions. A model joining mid-thread receives the conversation so far.

  4. A wallet is all the ID you need

    Try it with nothing at all. Connect Phantom for bigger limits. There’s no email, password or sign-up form.

Channels you can tune in

How it works

Three steps to a reading.

01

Choose your channels

Select up to four models for the question in hand. Premium ones need $TETRALYN or your own API key.

02

Send the question

Write it, paste a link, or attach a chart or a whitepaper. Each card fills in as its model finishes.

03

Take the reading

Request the verdict: one model chairs the comparison, merges the answers, scores agreement and lists each disagreement.

The reading

A number, with its reasons.

Agreement is scored from 0 to 100 and every disagreement is itemised, so you can see how much weight an answer will bear.

85%agreement
Consensus reached

70–100% agreement

Instruments

What’s on the bench.

Common sources

With Web on, all models read one fresh set of results and cite them as [1], [2]. Pasted links get read as well.

Files as input

Attach up to four images, PDFs or text files. A model without vision says it couldn’t see the image.

Memory across a thread

A newly added model catches up on the thread. Memory keeps it on your account if you switch it on.

Publish a reading

Share a public link and a generated image of the answers and the verdict.

Your keys, your billing

Connect OpenAI, Anthropic, Google and more to lift the daily limit. Keys are held in your browser.

Telegram bot

Message the bot a chart or a question. One reply carries all the answers and the verdict.

The token

$TETRALYN

Hold it to widen your limits.

Holding more raises your daily limits and opens GPT-6 Astra, Claude Opus 5, Gemini 3.1 Pro, Grok 4.6. There’s no staking or lock-up, and tokens remain in your wallet.

Contract address Dropping soon

Use only the contract address shown on this page. $TETRALYN is a volatile crypto asset, and this site offers no financial advice.

TierYou holdPremium / dayStandard / day
AnonNo login–2 questions
Probe Wallet login – 10
SignalUnlocks PRO 3M $TETRALYN 10 100
Array 10M $TETRALYN 40 300
Spectrum 25M $TETRALYN 100 1000

Burn for credits

10K $TETRALYN = 1 credit. Credits apply once the daily limit is used, don’t expire, and also work in the API. Burns are verified on-chain and can’t be reversed.

Open dashboard
0$TETRALYN burned
0burns
0burners
  • No burns yet.

Developers

Agreement as an API response.

One request returns all answers, the agreement score and the disagreements. Use the number in your logic: continue above 70, hand off to a person below 40.

  • Verdict mode, web search and file input
  • Standard answers cost 1 credit, premium answers 5
  • Rate limit: 60 requests per minute, per key
Read the docsGet an API key
~/agent
$ 
200 OK
{
  "consensus": {
    "agreement": 78,
    "verdict": "reached",
    "consensus": "Not yet. Measure first…",
    "disagreements": [{ "topic": "Team experience", … }]
  }
}

Agent tooling

Give your agent a reading before it acts.

Live

MCP server

Install Tetralyn in Claude, Cursor or any MCP client. The ask_models tool returns the verdict, the agreement score and the disagreements.

Claude Code
claude mcp add --transport http tetralyn https://tetralyn.fun/mcp \
  --header "Authorization: Bearer $API_KEY"
MCP setup guide
Soon

x402 per-call payment

No account or API key required. A request gets HTTP 402 with a price, the agent pays in USDC on Solana, then retries. Any unused amount is credited to the paying wallet.

1POST /v1/ask2402 · price3pay USDC4200 · reading
x402 details

Privacy

What gets recorded, and what doesn’t.

A plain account of the data involved and who handles your questions.

IdentifierWith a wallet, your public address. Without one, your IP address, used only to count free questions.
Login by signatureSigning a message logs you in. It’s free and isn’t a transaction, so it cannot move funds.
History is opt-inThreads are saved only when Memory is on. Turning it off deletes the saved thread.
Provider keysStored only in your browser. For each request they pass through our server, which doesn’t save them.
Who handles a questionIt travels through Tetralyn’s server to your chosen models, and each provider applies its own policies. Tetralyn trains no models. With Web on, a search provider sees the search as well.
Sharing is a choiceA question is public only after you create a share link. Anyone with that link can read it.

FAQ

Before you take a reading.

What is Tetralyn for?

Comparing AI answers. It sends a question to up to 4 models at once, shows the answers together and, on request, scores their agreement in a verdict.

Which models does it support?

GPT-6 Astra, Claude Opus 5, Gemini 3.1 Pro, Grok 4.6, DeepSeek V4 Pro, Mistral Medium 3.5, Cohere Command A+, Llama 4 Maverick, Qwen 3.8, Kimi K3, GLM-5.3. You can run up to 4 per question. New chats begin with two selected, and premium models need $TETRALYN or your own API key.

Is an account needed?

No. You get 2 free questions a day on standard models with no login. Connect a Phantom wallet for more; there’s no email or password.

When do I need $TETRALYN?

Only for premium models and higher daily limits. Standard models don’t need it, and your own API keys open premium models with no tokens held.

How is the score produced?

When the answers are in, press Get the verdict. One model chairs: it merges the answers, scores agreement (70% or more is consensus, 40–69% partial, under 40 no consensus) and lists each disagreement with the models behind it.

How are $TETRALYN tiers set?
  • No login: 2 free questions a day on standard models.
  • With Phantom: sign a free message (not a transaction) and your $TETRALYN balance is read.
  • 3M+ $TETRALYN (Signal): 10 premium and 100 standard answers daily.
  • 10M+ $TETRALYN (Array): 40 premium and 300 standard answers daily.
  • 25M+ $TETRALYN (Spectrum): 100 premium and 1000 standard answers daily.
  • Nothing is staked or locked. Limits reset each day at 00:00 UTC.
How do credits work?

Credits extend you past the daily limit at 1 per standard and 5 per premium answer. Burn $TETRALYN in the dashboard to get them (10K tokens = 1 credit). Burns are verified on-chain; credits never expire and work in the API.

Does it read the web?

Yes. Switch on Web and each model receives the same fresh sources, cited as [1], [2]. Pasted links are read automatically. A web search uses one extra answer and needs a connected wallet.

Where do my questions go?

Through Tetralyn’s server to your selected models, under each provider’s own policies. Chats are stored only with Memory on. Leave out anything you wouldn’t give those providers.

Does it accept my own API keys?

Yes. With a wallet connected, open Provider keys and add OpenAI, Anthropic, Google, xAI, DeepSeek, Mistral, Cohere or Hugging Face keys. There’s no daily limit, the provider bills you, and keys remain in your browser.

Why do models disagree?

Different training leads to different errors. A high score is a useful indicator rather than proof, so verify anything important.

Is $TETRALYN meant as an investment?

No. It’s a utility token for premium models, higher limits and credits. It confers no ownership or revenue share and comes with no promise of value. Crypto is volatile, so never risk more than you can afford to lose.

Take your first reading.

Open Tetralyn