MCP TrustOps for agent platform teams

Rank and route agents to MCP tools with fresh trust scores, policy exports, and runtime decisioning.

Audience

Agent platform teams

Primary API

/v1/decide

Policy exports

JSON, Rego, YAML, CI

Use cases

Choose the safest MCP server for a task before an agent calls it.
Route read-only work automatically and require approval for write actions.
Keep route decisions fresh as schemas, auth, and risk findings change.

Proof points

Agent Permissions Gateway returns allow, deny, or approval-required decisions for one agent/tool/user context.
Confidence-weighted scoring suppresses stale evidence by default.
MCP tools expose search, fetch, route_task, decide_agent_call, create_mandate, get_evidence_pack, export_policy, and governance options.

Sample flow

Flow

task -> route decision -> approval gate

Sample output

{
  "task": "search CRM and update opportunity stage",
  "recommended_server": "crm/example",
  "allowed_tools": ["search_accounts", "get_opportunity"],
  "blocked_tools": ["update_opportunity"],
  "approval_required": true,
  "reason": "write action detected"
}

Related buyer paths

Enterprise AI admins MCP publishers Security teams MCP marketplaces