AgentTrust MCP¶
Remote MCP server (Streamable HTTP) - AgentTrust sits between finding an agent and calling it. An endpoint URL goes in, the evidence AgentTrust has already observed for that exact URL is read, and a machine-readable trust decision comes out, all without contacting the endpoint itself.
Server type: Remote (Streamable HTTP)
Auth: See getagenttrust.com (key per account)
Endpoint: https://getagenttrust.com/api/mcp
Tools: check_agent_trust (returns a trustDecision object)
Pricing: See getagenttrust.com
Category: Business Operations
Built by: AgentTrust (getagenttrust.com)
Why This Matters for Operators¶
As agents start calling other agents, the missing step is the one a human does instinctively: deciding whether the counterpart is worth trusting before the call. AgentTrust fills that gap with evidence it has already collected about an endpoint, so the decision is based on observed history rather than a hopeful first request.
The differentiator is that it answers without touching the endpoint. A lookup reads stored evidence, so checking trust cannot itself trigger a slow, hostile or broken counterpart. The decision rule is deterministic and published: recommended when the endpoint is healthy and its current score is 50 or above, with confidence banded high at 90-plus and medium at 50-plus when ownership is verified, and low otherwise.
Tools & Capabilities¶
| Tool | Purpose |
|---|---|
check_agent_trust |
Takes an endpointUrl, reads the evidence AgentTrust has observed for that URL, and returns a trustDecision with health, reliability score, evidence freshness and endpoint ownership |
The returned decision carries recommended plus confidence and reasons, so a caller can act on the recommendation or apply its own policy to the raw signals. AgentTrust returns what it has observed; it is explicitly not a community rating and not a security guarantee.
Installation¶
npx add-mcp 'https://getagenttrust.com/api/mcp'
The same endpoint installs into Claude Code, Codex, Cursor and other clients through the standard MCP add flow.
Configuration¶
{
"mcpServers": {
"agenttrust": {
"type": "http",
"url": "https://getagenttrust.com/api/mcp"
}
}
}
Check the AgentTrust documentation for the authentication scheme on your account. A public test agent is available to try a lookup before wiring a real endpoint at https://getagenttrust.com/check-agent-trust?endpointUrl=https://agenttrust-umber.vercel.app/api/test-agent.
Business Relevance¶
- Platform teams building agent-to-agent workflows gate a call on a trust decision instead of invoking every discovered endpoint blind.
- Operators running multi-agent orchestrations screen a newly discovered endpoint for health and reliability before assigning it work.
- Security and risk teams add an evidence-based pre-call check in front of third-party agents, complementing their own policy.
- Marketplace and registry operators attach a trust signal to listed agents so consumers see observed behavior, not just a description.
- Anyone auditing an agent supply chain records the trust decision that preceded each external agent call.
Integration with CorpusIQ¶
CorpusIQ is the connector layer that lets AI assistants read the systems a business already runs on - accounting, payments, CRM, email, files and analytics - and answer with the source named. It is read-only by design: it retrieves data and never writes to, or acts in, a connected system, which is what makes it safe to point at live accounts.
This page is part of the CorpusIQ directory, the reference we maintain for the MCP server and agent-skill ecosystem. Run alongside CorpusIQ, a third-party server like this one can be combined with a company's own business data in a single answer.