Evidence-based grading for Solana tokens
Measuring whether a token’s utility is real.
Paste a contract address. Utilitest collects on-chain, web and code evidence, checks every utility claim against it, and assigns a tier from S to F.
Works on any SPL token. Tuned for pump.fun launches.
- Product
- 18
- Usage
- 06
- Token role
- 30
- Verifiability
- 52
- “AI agent is live”Contradicted
- “Staking launches soon”Unproven
- Mint & freeze authority revokedVerified
Method
A claim on a website is a hypothesis. Utilitest treats it like one: gather independent evidence, test the claim, and only then score.
-
01
Collect
Mint and freeze authority, supply and holder concentration from Solana RPC. Metadata, market and pool data. The project website and its docs, app and token pages. GitHub activity. Domain registration date.
-
02
Extract
Concrete utility claims are pulled from the token description and site copy: what the product does, who uses it, and what the token is for.
-
03
Verify
Each claim is checked against the collected evidence and labelled verified, partial, planned, unproven or contradicted, with the source cited. Promises tied to a future milestone are marked planned and listed under “watch for”, not counted against a new token.
-
04
Score
Four dimensions are scored 0–100 and weighted into a single score, which maps to a tier. Contradictions cap the result regardless of other scores.
Project-written text is treated as untrusted input: it can be the subject of a claim, never the evidence for one.
Rubric
Every token is scored on the same four dimensions. Weights reflect what matters most for utility: a product that works.
| Dimension | Weight | Counts as evidence | Does not count |
|---|---|---|---|
| ProductDoes it work today? | 35% | A live app or bot, a repository with real code, a working demo | Landing pages, roadmaps, “launching soon” |
| Token roleWhy does it need a token? | 25% | Documented tiers, staking, fees, buyback or burn, scoped governance; scored higher once seen working on-chain | “Ecosystem token”, “for the community” |
| UsageIs anyone using it? | 20% | Commit activity, integrations, observable product use | Trading volume, holder counts, followers |
| VerifiabilityCan claims be checked? | 20% | Docs, open source, consistent links, revoked authorities | Unlinked claims, anonymous screenshots |
Tokens under 14 days old can’t have a usage history yet, so usage is not scored for them: product counts 45%, token role 35%, verifiability 20%.
- S
- Shipped product with real users; the token is essential to it.
- A
- Working product with a clear, specific token role.
- B
- Product works, but the token’s role is thin, or vice versa.
- C
- Partial evidence. Something exists; much is unverified.
- D
- Mostly promises, or a claim contradicted by evidence.
- F
- No utility claimed, or no evidence for any of it.
Your tier list
Every token you evaluate is placed here automatically. Drag to re-rank, then export.
Stored locally in this browser.
$UTEST
A utility checker should pass its own test. The token has one job: access. Evaluations stay free to try; holding $UTEST removes the limits.
Not deployed. The address will be published here first.
Launch and contract address are announced on X first: @UtilitestDotFun
Limitations & questions
An evaluation is only as good as what can be observed from outside. Here is what that means in practice.
Does a high tier mean a token is a good investment?
No. Utilitest measures evidence of utility. It does not model price, liquidity risk or returns, and a token can trade well with an F or poorly with an A.
Who performs the evaluation?
The collection step is deterministic code. Claim verification and scoring are performed by a language model reading the full evidence pack under a fixed rubric. When that model is unavailable, a stricter rules-only evaluation runs instead, and the report labels which method produced it.
What can it not see?
Products behind logins, private repositories, X accounts, and activity that leaves no public trace. These show up as unproven, not as failures, and every report lists the sources it actually opened.
Why does a meme token get an F?
Because it makes no utility claims, so there is nothing to verify. The report labels it a pure meme. That is a description, not an accusation.
Can a project improve its tier?
Yes, by shipping. Each evaluation runs on fresh evidence. Publishing the app, linking the repository and documenting the token’s role all change the inputs.