Veridict

Know if you can trust your MCP server.

Veridict is a lightweight middleware that logs MCP tool executions and provides instant trust judgments. Add it to any MCP server in 2 lines.

Install

npm install veridict

Usage

import { McpServer } from "@modelcontextprotocol/sdk/server/mcp.js";
import { withVeridict } from "veridict";               // ← add this

const server = new McpServer({ name: "my-server", version: "1.0.0" });

// ... register your tools as usual ...

withVeridict(server, { name: "my-server" });            // ← add this
await server.connect(transport);

That's it. Every tool call is now logged and two new tools are available:

• veridict_stats — execution statistics for the server
• veridict_can_i_trust — trust judgment (YES / CAUTION / NO)

CLI

npx veridict                       # Show all tracked servers
npx veridict stats my-server       # Detailed stats
npx veridict trust my-server       # Trust judgment

How it works

1. Logs every tool execution — input hash, output hash, success/fail, latency
2. Stores locally — SQLite at ~/.veridict/executions.db
3. Judges trust — based on success rate over execution history

can_I_trust("my-server")?

  ✅ TRUSTWORTHY
  Confidence: 99.2%
  Reason: success_rate 99.2% over 1247 executions

Trust Judgment Logic

Options

withVeridict(server, {
  name: "my-server",            // Required: server identifier
  instanceId: "prod-1",         // Optional: distinguish instances
  dbPath: "./my-logs.db",       // Optional: custom DB path
  minExecutions: 20,            // Optional: min data for judgment (default: 10)
});

Future

Veridict is the trust layer for the AI agent economy. Current version stores logs locally. Future versions will support:

• On-chain audit trails (XRPL)
• Cross-agent trust verification
• Replay verification
• Enterprise compliance reporting

License

MIT