AgentForge Trust · v1 live

The audit layer for MCP servers.

Enterprise AI agents shouldn't trust random MCP servers. AgentForge runs a five-dimensional audit — security, code health, behavior, community, EU compliance — on every server in the catalog. Integrate once, filter agent tool discovery by trust score forever.

Explore audited catalog Install MCP tool

23,402

MCP servers indexed

3,652

Audited (16%)

Avg top-100 score

Five dimensions, one score

Security Scan

30% weight

Snyk, GitHub Advisory, secret scanning, dependency CVEs.

Code Health

20% weight

Commit recency, open-issue ratio, license presence, archived state.

Behavioral Audit

20% weight

Claude-powered source review — what does it actually do vs. README.

Community Trust

15% weight

Stars (log-scaled), forks, author reputation, organization bonus.

EU Compliance

15% weight

GDPR, AI Act readiness, data residency disclosure.

Top-audited servers

Server	Score	Badges
	100	actively maintainedcommunity trustedlicensed · apache-2.0
	100	actively maintainedcommunity trustedlicensed · apache-2.0
	100	actively maintainedcommunity trustedlicensed · apache-2.0
	100	actively maintainedcommunity trusted

Integrate in one line

Install the @agentforge/trust-mcp server in your agent. It exposes four tools: check_trust, evaluate_policy, list_trusted, recommend.

{
  "mcpServers": {
    "agentforge-trust": {
      "command": "npx",
      "args": ["-y", "@agentforge/trust-mcp"]
    }
  }
}

Or hit the public API directly: GET /api/v1/trust?slug=… · POST /api/v1/trust/evaluate · GET /api/v1/trust/list