MCP Service Trust Evaluator

Check if an MCP server is safe before you spend money on it

evaluate_service checks any MCP server's reachability, governance declarations, tool definition quality, and audit endpoints, returning a trust score from 0 to 100 with a recommendation. Pass it a server URL and get back a clear verdict: PROCEED, CAUTION, HIGH RISK, or DO NOT TRANSACT.

Free to use. No API key required. Works from any MCP client.

What It Checks

Each evaluation runs six checks against the target MCP server. The trust score is a weighted composite of all six.

• Reachability -- can the MCP endpoint be reached, and how fast does it respond
• Governance declarations -- does the server publish governance invariants that describe its operating rules
• Tool definition quality -- are tool names, descriptions, input schemas, and annotations well-formed and complete
• Audit endpoints -- does the server expose an audit log or transparency endpoint for accountability
• agent.json -- does the server publish a /.well-known/agent.json discovery file
• HTTPS -- is the connection encrypted end to end

Trust Score Recommendations

The trust score maps to a plain-language recommendation so agents and developers can make fast decisions.

• PROCEED (75-100) -- server meets governance standards, safe to transact
• PROCEED WITH CAUTION (50-74) -- some governance gaps, review before authorizing payments
• HIGH RISK (25-49) -- significant governance concerns, not recommended for sensitive tasks
• DO NOT TRANSACT (0-24) -- server does not meet minimum trust standards

How to Use It

Connect to the Agentic Platform MCP endpoint, then call evaluate_service with the URL of the MCP server you want to evaluate.

Step 1: Connect

claude mcp add agentic-platform --transport http https://api.asalvocreative.com/mcp

{
  "mcpServers": {
    "agentic-platform": {
      "url": "https://api.asalvocreative.com/mcp"
    }
  }
}

npx -y @smithery/cli@latest install agentic-platform--andysalvo.run.tools --client claude

Step 2: Call evaluate_service

evaluate_service({
  server_url: "https://example.com/mcp",
  task_context: "I need to use this server for code generation"
})

Example Output

Service Trust Evaluation
========================
Target: https://example.com/mcp
Trust Score: 62/100
Recommendation: PROCEED WITH CAUTION

Checks:
  PASS  Server reachable (238ms response time)
  PASS  Valid MCP endpoint (12 tools found)
  PASS  HTTPS enabled
  FAIL  No /.well-known/agent.json found
  FAIL  No audit endpoint
  WARN  3 tools missing descriptions
  WARN  No governance declarations

Summary: Server is functional but lacks transparency.
         Review tool definitions before authorizing payments.

When to Use evaluate_service

Run an evaluation before connecting to any MCP server you have not used before. This is especially important before authorizing payments, sharing sensitive data, or delegating critical tasks to a third-party service.

AI agents can call evaluate_service programmatically as part of their workflow. Before an agent transacts with a new MCP service, it can check the trust score and only proceed if the score meets a threshold.

About the Agentic Platform

The Agentic Platform is an MCP server that provides free agent diagnostics and expert skill files. In addition to the service evaluator, it offers an MCP linter, agent health check, and cost estimator -- all free, no API key required.