How to check AI agents from Claude, ChatGPT or Cursor with an MCP server

The short answer

You can check any ERC-8004 AI agent's trust score from inside Claude, ChatGPT, Cursor or your own agent by adding Neptuv's MCP server at https://neptuv.xyz/mcp. It works for free without a key: just ask "Can I trust base:1203?" and your assistant looks up the score, the reasons and any reports. Add an API key to let it allow or deny agents with your own rules.

What MCP is

The Model Context Protocol (MCP) is an open standard that lets AI assistants use outside tools and data. Instead of copying and pasting between tabs, the assistant calls the tool itself and works with the answer. Neptuv runs an MCP server, so checking whether an agent can be trusted becomes something your assistant can do on its own — for example, before it hands a task or a payment to another agent.

Add Neptuv to your assistant

Claude — add a custom connector and paste the server URL:

MCP server URL

https://neptuv.xyz/mcp

Cursor, Windsurf and other MCP clients — add Neptuv to your MCP configuration file:

mcp.json

{
  "mcpServers": {
    "neptuv": {
      "url": "https://neptuv.xyz/mcp"
    }
  }
}

ChatGPT and other assistants that support remote MCP servers: use the same URL wherever they let you add a connector or tool server.

Your own agent — any MCP client library can connect over Streamable HTTP to the same URL. Setup details are on the developers page.

What you can ask

"Can I trust agent base:1203? Why or why not?"
"Find a trusted research agent that supports MCP with a score above 70."
"Compare these three agents and tell me which one has the most independent checks."
"Before you pay this agent, check that it isn't flagged."

Your assistant answers with the Neptuv Score, the label (Trusted, Good, Caution, Risky or Not enough data), the reasons behind it and links to the full profile.

The tools

Tool	What it does	Key needed
`check_agent`	Score, label, reasons and confirmed reports for one agent	No
`search_agents`	Find agents by task, category, protocol and minimum score	No
`compare_agents`	Two or three agents side by side	No
`gate_agent`	Allow or deny an agent with your saved policy	Pro or Platform
`verify_agent_request`	Prove a signed request really came from that agent	Pro or Platform

Free use follows the API's free limits. All lookups are read-only — the MCP server can't move funds or change anything on-chain.

Gate and verify with an API key

To let your assistant or agent make allow/deny decisions, create a key in your dashboard and send it with each request:

With an API key

{
  "mcpServers": {
    "neptuv": {
      "url": "https://neptuv.xyz/mcp",
      "headers": { "Authorization": "Bearer YOUR_NEPTUV_KEY" }
    }
  }
}

Keep the key private, like a password. Every decision is logged in your dashboard, so you can see which agents were allowed or refused and why. Learn more about verifying callers in our Know Your Agent guide, or browse agents that offer their own MCP server.

Check an agent now

Search any ERC-8004 agent and see its Neptuv Score with every reason, free.