Server data from the Official MCP Registry
Know if you can trust your MCP server. Trust judgment for AI agents.
Know if you can trust your MCP server. Trust judgment for AI agents.
Valid MCP server (2 strong, 2 medium validity signals). 1 code issue detected. 3 known CVEs in dependencies (0 critical, 3 high severity) Package registry verified. Imported from the Official MCP Registry. 1 finding(s) downgraded by scanner intelligence.
11 files analyzed · 5 issues found
Security scores are indicators to help you make informed decisions, not guarantees. Always review permissions before connecting any MCP server.
This plugin requests these system permissions. Most are normal for its category.
Add this to your MCP configuration file:
{
"mcpServers": {
"io-github-xkumakichi-veridict": {
"args": [
"-y",
"veridict"
],
"command": "npx"
}
}
}From the project's GitHub README.
Know if you can trust your MCP server.
Agents call tools. Some fail silently. Veridict gives you a signal.
Part of the emerging Agent Trust Stack (Runtime Verification Layer).
npm install veridict
import { withVeridict } from "veridict"; // 1. import
// ... register your tools as usual ...
withVeridict(server, { name: "my-server" }); // 2. wrap
await server.connect(transport);
Done. Every tool call is now logged. Two new tools are automatically added:
veridict_stats — execution statisticsveridict_can_i_trust — trust judgment (YES / CAUTION / NO){
"verdict": "yes",
"score": 0.97,
"reason": "reliable",
"recentSuccessRate": 0.98,
"recentExecutions": 85,
"failureBreakdown": { "timeout": 2, "error": 1 },
"stats": {
"totalExecutions": 1247,
"successRate": 0.992,
"avgLatencyMs": 142
}
}
Two-layer design:
verdict + score + reason — all an agent needsstats + failureBreakdown + recentSuccessRate — dig deeper when neededtimeout, error, validation, or unknownscore field (0-1) replaces confidence for clearer semantics"elevated timeout rate", "reliable", "high error rate"See executions in real-time:
withVeridict(server, { name: "my-server", verbose: true });
Output (stderr):
[veridict] monitoring "my-server"
[veridict] search_docs ok 120ms
[veridict] create_item ok 85ms
[veridict] fetch_data FAIL [timeout] 30201ms — timeout
[veridict] search_docs ok 94ms
npx veridict # Show all tracked servers
npx veridict stats my-server # Detailed stats
npx veridict trust my-server # Trust judgment
| Blended Rate | Verdict | Meaning |
|---|---|---|
| >= 95% | yes | Trustworthy |
| >= 80% | caution | Some failures detected |
| < 80% | no | High failure rate |
| < 10 executions | unknown | Insufficient data |
Blended rate = recent 7-day success rate (70% weight) + all-time rate (30% weight).
Recent performance matters more than history.
| Type | Detected when |
|---|---|
timeout | Latency > 30s, or error contains "timeout"/"ETIMEDOUT" |
validation | Error contains "valid"/"schema"/"parse" |
error | All other failures |
unknown | Legacy data (pre-v0.2.0) |
withVeridict(server, {
name: "my-server", // Required: server identifier
instanceId: "prod-1", // Optional: distinguish instances
dbPath: "./my-logs.db", // Optional: custom DB path (default: ~/.veridict/executions.db)
minExecutions: 20, // Optional: min data for judgment (default: 10)
verbose: true, // Optional: log to stderr (default: false)
});
Combine static analysis with runtime monitoring for full-stack trust:
# Step 1: Static scan with MCP Trust Kit (v0.5.0+)
npx mcp-trust-kit scan --json-out layer1-baseline.json --cmd node my-server.js
import { withVeridict, parseLayer1Report } from "veridict";
import report from "./layer1-baseline.json";
withVeridict(server, {
name: "my-server",
baseline: parseLayer1Report(report), // Layer 1 → Layer 2
});
What this does:
dangerous_fs_write) are factored into trust judgmentscan_timestamp is preserved for Layer 3 temporal decay logicbaseline.raw for cross-org consumersPart of the Agent Trust Stack:
Layer 1: MCP Trust Kit (pre-deploy) → "Is this server safe to run?"
Layer 2: Veridict (runtime) → "Is this server actually reliable?"
Layer 3: SATP/XAIP (cross-org) → "Should I trust this across boundaries?"
~/.veridict/executions.db)If you're building MCP servers or agents, I'd love your feedback. Try Veridict and tell me what breaks.
MIT
Be the first to review this server!
by Modelcontextprotocol · Developer Tools
Read, search, and manipulate Git repositories programmatically
by Toleno · Developer Tools
Toleno Network MCP Server — Manage your Toleno mining account with Claude AI using natural language.
by mcp-marketplace · Developer Tools
Create, build, and publish Python MCP servers to PyPI — conversationally.