FIRAX AI speaks the Model Context Protocol (MCP), the open standard many business applications now support. Anything with an MCP server can be connected by pasting its endpoint; anything without one can still be brought in.
Before you start
- The application's administrator, and agreement on what FIRAX AI may read and do.
- Preferably a dedicated, least-privilege account in the application for the connection.
Steps
If the vendor offers an MCP server
Ask the vendor or your partner for the endpoint, then in FIRAX AI open Gatekeepers → Any MCP server, paste it and sign in. Grant named tools only to start.
If the application has an API but no MCP server
Put an MCP server in front of the API — an open-source one for that product, or one the FIRAX team builds for you — and connect that endpoint the same way.
If there is no API
The FIRAX team can build a custom FIRAX AI connector for the system (a Gatekeeper), for example over its export files or database views, deployed in FIRAX's Cloudflare account.
Decide what needs approval
Read-only tools run straight away; every tool that creates or changes something waits for a person to approve it. Keep writes out of the grant until you need them.
Good to know
- Prefer a read-only service account for reporting use cases.
- Every call is logged on the connection — review it in FIRAX AI, and in the application's own audit log.
- Talk to us at hello@firax.ai about a connector for a system that is not listed.