Independent trust verification for every MCP your agents install or call.
Every M8ven AI Tool Trust verdict is built on what the MCP actually does — not on what it claims about itself. MCPs that pass earn the M8ven Verified badge and appear in the MCP Trust Index.
What you get
Static analysis tells you what an MCP claims. M8ven AI Tool Trust tells you what it actually does.
Every check returns an explainable result — ok, warning, or fail — with the underlying evidence attached. No star ratings, no vibes. Queryable by GitHub URL or MCP name.
Not just a score on the whole MCP. Each tool the MCP exposes gets its own observed outcome, so you know exactly which tools were verified and which ones the MCP won't even let us test.
We don't trust what an MCP says about itself. Our verdicts come from what the MCP actually did when we exercised it — which hosts it contacted, which files it touched, which processes it spawned.
Drop M8ven AI Tool Trust into your MCP directory as a filter, a badge, or a full report. REST + webhook + MCP-native. Your users see the M8ven Verified checkmark, you skip the verification infrastructure.
For MCP registries
Drop M8ven AI Tool Trust into your listing page as a filter, a badge, or a full report. Your users see the checkmark, you skip the verification infrastructure.
For AI platforms
Call M8ven AI Tool Trust before your agent imports a new MCP. If the verdict is red, surface it. If it's green, proceed. Fewer supply chain surprises, no new infra to run.
For MCP developers
Submit your MCP, pass verification, get the M8ven Verified badge for your README and your listing. Users trust what we vouched for.
AT2 is in public beta
We're actively calibrating the sandbox and expanding coverage. The API is stable enough to build against, and the directory at /verified reflects real sandbox runs. But until we ship enough of the agentic stack for AT2 to pay its own way, it runs free — gated only by a short waitlist so we can keep the load manageable.
Get on the waitlist