# Neptuv

> Neptuv is a trust score for AI agents. It reads public ERC-8004 data and turns it into a score from 0 to 100, with every reason shown.

- The Neptuv Score (method v1.1) is a number from 0 to 100 built from four signal groups: Reputation (35 points), Validation (25), Identity (20) and Track record (20), minus penalties for risk flags.
- Labels: Trusted (80–100), Good (60–79), Caution (40–59), Risky (0–39). "Not enough data" means fewer than three distinct reviewers and validators in total. "Flagged" means the owner or agent wallet is on a public sanctions list (score 0).
- Feedback loses half its weight every 120 days. Each reviewer counts once. Brand-new wallets that rated only one agent count for 40%, and so do reviewers who only ever rate one owner's agents. Ratings traded between owners who review each other are ignored and cost 10 points. A 24-hour burst of feedback from brand-new wallets costs 15 points. Owners running other agents with active risk flags lose 10 points on each agent.
- Data source: the ERC-8004 Identity, Reputation and Validation registries, deployed at the same addresses on every mainnet: Identity 0x8004A169FB4a3325136EB29fA0ceB6D2e539a432, Reputation 0x8004BAa17C55a88189AE136b182e5fdA19dE9b63, Validation 0x8004Cc8439f36fd5F9F049D9fF86523Df6dAAB58.
- Neptuv only reads public data. It never holds or moves user funds and never asks for private keys or seed phrases.
- Plans: Free (search, scores, 100 API calls a day), Builder $50/month, Pro $100/month, Platform $500/month. Paid in USDC on Base, or USDC, USDT or ETH on Ethereum. A year costs the same as 10 months.
- Owners cannot pay to change a score; paid plans unlock tools (badge, alerts, API), not points.
- Know Your Agent: agents sign each request with their ERC-8004 wallet (headers Neptuv-Agent, -Timestamp, -Nonce, -Signature; EIP-191). Platforms call POST /api/v1/verify to prove who is calling and get allow/deny from their saved policy. Replays and stale timestamps (over 5 minutes) are refused.
- Incident network: Pro and Platform workspaces report misbehaving agents; reports stay private until Neptuv confirms them; a confirmed report adds a public risk flag (−25) shown with the owner's reply.
- Integrations: REST API, OpenAPI spec (/openapi.json), MCP server (/mcp), Python and JavaScript SDKs, signed webhooks, Telegram bot, embeddable widget and EIP-712 attestations verifiable on-chain.

## Core pages
- [How the Neptuv Score works](https://neptuv.xyz/methodology): the full method with every weight and threshold
- [FAQ](https://neptuv.xyz/faq): short answers about the score, ERC-8004, fake reviews, disputes and payments
- [Explore agents](https://neptuv.xyz/agents): every indexed ERC-8004 agent with its score and reasons
- [Know Your Agent](https://neptuv.xyz/verify): verify which AI agent is calling your API, with a live playground
- [API documentation](https://neptuv.xyz/docs): REST endpoints for scores, search, verify, policies and the allow/deny gate
- [Developers](https://neptuv.xyz/developers): SDKs, MCP server, webhooks, widget and attestations
- [Incident network](https://neptuv.xyz/network): how businesses report bad agents and how reports are reviewed
- [Pricing](https://neptuv.xyz/pricing): plans and how crypto payments work
- [OpenAPI spec](https://neptuv.xyz/openapi.json): machine-readable API description
- [MCP server](https://neptuv.xyz/mcp): Model Context Protocol endpoint (Streamable HTTP) for AI assistants

## Guides
- [What is an AI agent trust score? How on-chain reputation works](https://neptuv.xyz/learn/ai-agent-trust-score): What a trust score for AI agents measures, how it stops fake reviews, how to read one and where it falls short.
- [What is ERC-8004? The standard behind AI agent identity](https://neptuv.xyz/learn/what-is-erc-8004): How the Identity, Reputation and Validation registries work, where they are deployed, and what an agent registration file contains.
- [AI agent scams: 9 warning signs before you trust an agent](https://neptuv.xyz/learn/ai-agent-scams): How AI agent scams work, the warning signs to check before an agent touches your money or data, and what to do if you've been scammed.
- [How to check whether an AI agent can be trusted](https://neptuv.xyz/learn/how-to-check-an-ai-agent): A practical checklist for platforms and buyers: identity, feedback, independent checks, track record and red flags.
- [x402 explained: how AI agents pay for APIs, and how to know who's paying](https://neptuv.xyz/learn/x402-ai-agent-payments): How the x402 standard uses HTTP 402 and stablecoins for pay-per-request payments, and why you still need to know which agent is paying.
- [Know Your Agent (KYA): how to verify which AI agent is calling your API](https://neptuv.xyz/learn/know-your-agent): Why API keys and user agents can't identify AI agents, how signed requests tied to an ERC-8004 identity work, and how to add an allow/deny gate in front of your platform.
- [How to check AI agents from Claude, ChatGPT or Cursor with an MCP server](https://neptuv.xyz/learn/check-ai-agents-mcp): Add Neptuv's MCP server to your AI assistant and check any agent's trust score, compare agents and gate them with your own rules — without leaving the chat.

## Optional
- [Full text for language models](https://neptuv.xyz/llms-full.txt)
- [State of ERC-8004 agents](https://neptuv.xyz/report): live statistics