AI AGENT STORE / SERVICES

APP INTEGRATIONS · FOR AI AGENTS

Permit MCP Gateway

Route MCP tool calls through managed identity, authorization and audit controls.

Permit MCP Gateway sits between an agent’s MCP client and external MCP servers. Administrators configure user and agent access, tool trust levels and audit logging so connected clients can use authorized tools through a shared gateway.

Your integration checklist

What you need from your owner
Create a gateway organization and host, connect the intended MCP servers, configure user and agent identities and assign tool access. Point clients at the host-specific gateway URL and complete the required sign-in.
Authentication
Use the gateway host’s configured authentication flow. Upstream server credentials and consent remain separate dependencies. A direct connection to the upstream server bypasses gateway decisions and logging.
Your first task
Connect one staging MCP server, run a permitted read and a deliberately denied call, then verify that the audit record identifies the intended caller and tool.
Billing & payment
The Community gateway plan requires no credit card and includes limited active identities, tenants, environments and audit retention. Humans and agent identities can count separately. OAuth proxy, longer retention and advanced deployment features have plan-specific charges.
Agent payment support
The limited Community gateway plan is documented as requiring no credit card. Paid features and upstream service usage require separate billing arrangements.
Operational limits
Trust levels are hierarchical, and tool-name classification is not proof of a tool’s behavior. Gateway policy governs calls routed through it; it does not replace authorization inside the upstream service. Feature availability differs between gateway and general Permit plans.

ORIGINAL AI AGENT STORE INTEGRATION PLAYBOOK

Check the actual tool behavior behind each access tier

Editorial recommendation · Reviewed 2026-11-10

When to choose this service

Choose the hosted gateway when a team needs shared MCP access administration and audit records across providers. Native service permissions may suffice for one trusted client and a small tool set.

Hosted identity sessions, policy enforcement and shared audit infrastructure provide an external access service. A prompt cannot grant another user’s consent or substitute for continuously operated authorization infrastructure.

Suggested workflow

  1. Inventory the intended tools and their actual effects, then assign the minimum tier that supports the task. Confirm the relevant plan features.
  2. Test allowed and denied calls through the gateway using the intended identity. Inspect the audit trail and check whether alternate direct connections remain available.
  3. Recheck permissions when upstream tools change, and export or retain audit evidence within the configured retention period.

Acceptance checks before real work

  • A read-like tool name does not guarantee a read-only effect.
  • Higher trust tiers include lower-tier permissions.
  • Calls made directly to an upstream server are outside the gateway path.

If the workflow fails

For unexpected denial, inspect caller identity, host configuration, granted tier and upstream consent. Resolve the specific mismatch instead of automatically assigning a broader tier.

Provider documentation

Explore services for adjacent needs

Continue building your toolkit