Connecting an AI Agent

To work with your rules from an AI client, you add your application's MCP server to the client once, as a connector, and sign in. From then on, the agent can use your rules in any conversation where the connector is turned on. This page explains what you need and the usual steps; the exact menu names differ from one AI client to another and change over time, so follow your client's own help for the details.

What you need:

  • The server's address, such as https://rules.yourcompany.com/mcp. Your application's administrator gives it to you.
  • Your usual sign-in for your application. The AI client sends you to your company's sign-in page; the agent never sees your password.
  • An AI client that supports MCP connectors, such as Claude on the web, on your desktop or on your phone, or an AI coding tool.

Adding the Connector in Claude

In company plans of Claude, an organization owner usually adds the connector once for everyone, and each person only connects to it:

  1. In Claude, open the connectors settings.
  2. Find your application's connector in the list your organization added, and click Connect.
  3. Your company's sign-in page opens. Sign in as you always do, including any second step your company uses, and approve the access.
  4. You return to Claude with the connector turned on.

In personal plans, you add the connector yourself: in the connectors settings, add a custom connector, give it a name, paste the server's address, and continue with the sign-in as above.

Claude connects from Anthropic's servers, not from your computer, so your application's server must be reachable from the internet, or at least from Anthropic. If the connection fails, ask your administrator.

Adding the Server to an AI Coding Tool

AI coding tools take the server's address in their settings or configuration file, under the name of an MCP server using HTTP. When the tool first connects, your company's sign-in page opens in your browser. Sign in, and the tool is connected.

Using the Connection

  1. Start a conversation and make sure the connector is turned on for it.
  2. Ask in plain language, for example "List the rules for orders and tell me which ones are live."
  3. The agent shows when it uses the connector. When it needs your confirmation, for example before changing an existing rule or running a test that calls your application's real methods, your client asks you; answer it, or tell the agent in the conversation.

The agent sees only what your permissions allow. If it says it cannot do something, it may be a permission you do not have, or one of the limits that apply to every agent. See What Agents Can Do.

Disconnecting

To stop an agent from using your rules, turn the connector off in a conversation, or disconnect it in your client's connectors settings. Your administrator can also remove the connector for the whole organization.

Ask on Stack Overflow - Yes, it's still active
Ask in GitHub Discussions

p102

l097 --

l102

p101

×