Introduction to the MCP Gateway
Updated on 3 November 2026.
The MCP Gateway sits between your AI apps and the MCP servers they call. Users connect their AI app to a Quilr gateway URL instead of the server itself, and every tool call is checked against your tool rules and guardrails, then logged with the user, the agent and the result.
Key concepts
New servers start with their tools turned off, so nothing is reachable until you enable it.
The MCP Gateway page
Go to Settings > AI Gateway > MCP Gateway.
Click a row on a tile to filter the server list or open the matching view.
Below the banners, Elsewhere links to Inventory & analytics, the User dashboard and this Documentation. The server list has All, Local, Remote and API tabs, Search servers and a Filter for state, auth, source and runtime.
The server card
Each configured server has a card.
A server waiting for an administrator OAuth sign-in shows Connect OAuth on its card.
Analytics and insights
Overall analytics opens a full page with Overview, Usage, Data & security, Reliability, and Servers & tools tabs. Choose the period, server scope and AI-client scope before comparing traffic or following a chart into its details.
See Analytics and insights for adoption, access maps, failure investigation and the difference between tool-list savings and smaller tool results.
The workspace
The server card's Inspect and Configure actions open operational views. Inspect > Logs opens that server's Activity; Configure opens its settings. These views are separate from full-page Insights. Select the server scope and period in the workspace; Settings needs exactly one server.
Settings sections
Sections vary by server. Edits are drafts until you click Save settings.
When the Policy Engine is on, Settings shows Policy Engine active and marks Tools, Guardrails, Token saving, Tool list and Group & User Rules with its icon. Those sections follow published policies instead of these settings. General and API tokens are still managed here.
Next steps
- Quick Start: add a server and make a first tool call.
- Analytics and insights: review adoption, security, failures and tool savings.
- Adding MCP Servers: remote servers, sign-in modes and the MCP Library.
- API to MCP: turn a REST API into tools.
- OneMCP: one URL for every MCP a user may use.
- Integration Guide: URLs, tokens and client examples.