Agents spend real money
With payment standards like x402, agents pay for services on their own. One bad agent can burn a budget or leak data before anyone notices.
Live on Ethereum & Base · ERC-8004
Know your agent. Trust, calculated.
The trust score for AI agents. Neptuv reads public ERC-8004 data — who rated an agent, who checked its work, how long it has existed — and turns it into one score from 0 to 100, with every reason shown.
1,240,319agents scored
140,560distinct reviewers
12independent checks
Why it matters
AI agents now book, buy and trade for people. Anyone can launch one in minutes. Nobody can easily tell which ones to trust.
Reviews are cheap to fake
New wallets cost nothing. Raw on-chain feedback can be pumped by a few hundred throwaway reviewers in an afternoon.
Raw data is unreadable
ERC-8004 records identity, feedback and validations on-chain. Useful, but scattered across chains and impossible to judge by eye.
The Neptuv Score
Four signals. One number. Every point explained.
Every agent gets a score from 0 to 100. Here is exactly where the points come from. The full method is public and versioned.
35 pts
Reputation
What clients say about the agent, weighted so fake reviews barely move it.
- Each reviewer counts once, however many times they post
- Recent feedback counts more: it loses half its weight every 120 days
- Brand-new wallets that rated only this agent count 40%
- Reviewers who only ever rate one owner's agents count 40%
- Feedback from the owner's own wallets is ignored
25 pts
Validation
Independent checkers that re-run or inspect the agent's work and post a result.
- Validator scores from 0 to 100, recorded on-chain
- Several validators agreeing beats one
- Pending requests are shown, never counted
20 pts
Identity
Who stands behind the agent and how much they have put on the line.
- Age of its on-chain identity (full points at 90 days)
- A complete registration file
- Its domain proven with a
.well-knownfile - An owner wallet with a real history
20 pts
Track record
How the agent has behaved over time.
- Months of active use
- Stable ownership (a recent transfer costs points)
- Endpoints that actually respond
−25 per report
Risk flags
Penalties that pull a score down when something looks wrong.
- A burst of feedback from brand-new wallets: −15
- Agents rating each other back (a rating ring): −10
- Each confirmed report: −25, and −10 on the owner's other agents
- Marked inactive by its owner: −10
- Owner or agent wallet on a sanctions list: score set to 0
For platforms
From a score to a decision, in one call.
A score tells you who to trust. Neptuv also proves which agent is calling, applies your rules, and keeps the receipts.
Verify who's calling
Agents sign each request with the wallet registered on ERC-8004. Neptuv checks the signature against the registry and blocks replays — impersonators get nowhere.
Know Your Agent →Decide with your rules
Minimum score, independent checks, identity age, open reports, allow and deny lists. Save policies once; every answer is logged for audits.
Policies →Prove it on-chain
Signed EIP-712 attestations let smart contracts require a minimum trust score — checked by a few lines of Solidity, no oracle needed.
Attestations →
ERC-8004
MCP
A2A
x402
Python
JavaScript
Solidity
Telegram
The network
One bad agent, every platform warned.
Businesses report agents that scam, spam or don't deliver. Neptuv reviews each report and hears the owner out; confirmed findings lower the score everywhere it's checked. The more platforms join, the safer every one of them gets.
Who it's for
For anyone who lets agents in.
Platforms and marketplaces
Decide which agents may buy from you, call your API or touch your data. Verify who's calling and get allow or deny from your own rules.
Know Your Agent →Agent builders
Claim your agent with a wallet signature, see exactly what holds your score back, and show a verified badge on your site.
Verified badge →Developers
REST API, one-file SDKs, an MCP server for AI assistants and signed webhooks. Free to start.
Developers →API
One call. A clear answer.
Ask for any ERC-8004 agent and get its score, label and reasons back in milliseconds. On Pro, the gate and verify endpoints turn your rules into a simple allow or deny.
Request
curl https://neptuv.xyz/api/v1/agents/base/1203/score \
-H "Authorization: Bearer nep_your_key"Response
{
"name": "Atlas Research",
"score": 86,
"label": "Trusted",
"reasons": {
"up": ["Rated by many different reviewers"]
}
}Read-only by design
Built so you never have to trust us with money.
Never touches funds
Neptuv only reads public data. It never holds, moves or asks for anyone's money, keys or seed phrase.
Sign in with a free signature
Prove you own a wallet without a transaction or gas fee. No email or password needed.
Open, versioned method
Every weight is published. When the method changes, the version number changes with it.
Scores can't be bought
Paid plans unlock tools, never points. Every agent is scored the same way.
Public corrections
Owners can dispute a score. Every correction is logged on the agent's page.
Built on an open standard
Data comes straight from the official ERC-8004 registries, the same on every chain.
Questions
The short answers.
What is Neptuv?
Neptuv is a trust score for AI agents. It reads the public ERC-8004 registries on Ethereum, Base and other chains and turns what it finds into one score from 0 to 100, with the reasons written in plain words. Businesses use it to decide whether to let an agent buy from them, call their API or access their data.
What is the Neptuv Score?
A number from 0 to 100 built from four groups of public signals: Reputation (up to 35 points), Validation (up to 25), Identity (up to 20) and Track record (up to 20). Points are taken away for risk flags. 80–100 is Trusted, 60–79 Good, 40–59 Caution and 0–39 Risky.
What is ERC-8004?
ERC-8004 ("Trustless Agents") is a draft Ethereum standard that gives AI agents an on-chain identity. It defines three registries: Identity (each agent is an NFT pointing to a registration file), Reputation (feedback from clients) and Validation (results from independent checkers). The registries are deployed at the same addresses on many chains.
How is the score calculated?
Neptuv adds up points from four signal groups and subtracts penalties. Feedback counts more when it comes from many different, established reviewers and when it is recent. Independent validations count more when several validators agree. The full method, with every weight, is published on the methodology page.
Know your
agent.
Search any ERC-8004 agent. Free, no account needed.