Agentradar MCP

What Bichev/agentradar-mcp MCP server does

Bichev/agentradar-mcp MCP server connects an MCP client to the AgentRadar API at https://api.vvpro.ai. It is a thin client wrapper rather than an independent scoring database or blockchain indexer. The package exposes 18 tools for evaluating AI agents and wallet addresses through trust scores, identity data, reputation signals, scam records, and attestations.

The main verification flow returns a composite score from 0 to 100, the individual signal scores, risk flags, and recommendations. Other tools support side-by-side comparison, batch verification of up to 10 addresses, cached score retrieval, human-readable score explanations, and embeddable trust badge URLs.

How it works

The server sends MCP tool requests to the public AgentRadar API. verify_agent runs the full six-signal assessment, while get_score retrieves the last computed score without rerunning the signals. list_signals provides the scoring signals, their weights, and descriptions, which can help an application explain how a result was formed.

Identity and ecosystem checks cover ERC-8004 registrations, ENS names, contract metadata, feedback history, attestation counts, community signals, and registered agent URIs. check_scam performs a focused lookup against the AgentRadar scam database, returning records such as severity, chain, source, and notes when available.

For permanent records, attest_agent scores an agent and writes an on-chain EAS attestation. Existing attestations can be listed or retrieved by transaction hash. These write and administrative operations have different access requirements from free lookups.

Setup and configuration

The package can be started directly with:

npx -y @agentradar/mcp

It is documented for Claude Desktop and Cursor, using an MCP configuration with the npx command and @agentradar/mcp package. The same package can also be built locally after cloning the repository, installing npm dependencies, running the build, and starting node dist/index.js.

AGENTRADAR_API_URL optionally overrides the default API base URL. AGENTRADAR_OPERATOR_KEY enables operator-only endpoints and bypasses x402 payment where the API permits it. The operator key is not required for the basic public setup described in the repository.

Tools and capabilities

Bichev/agentradar-mcp MCP server includes these capability groups:

• Verify one agent, compare two agents, or batch-check up to 10 addresses.
• Retrieve cached scores, explain score composition, and list scoring signals.
• Look up ERC-8004 identity, reputation, registered URI, metadata, and service endpoints.
• Check or report scam wallets.
• Create, list, and inspect on-chain EAS trust attestations.
• Check API health, retrieve dashboard statistics, and inspect pricing.
• Generate a trust badge URL for use in an interface or document.

Limitations and notes

The package depends on the AgentRadar API and its available on-chain data. A score is composed from the service’s six signals; the repository does not describe the signals as a substitute for independent transaction review. Batch verification is limited to 10 agents per call.

get_score is free, while verify_agent is listed at $0.005 per call and attest_agent at $5 per attestation when paid through x402. The repository states that these calls can be free with an operator key. Administrative endpoints, including statistics and several attestation or reporting operations, require that key. The project is released under the MIT license.