Using Magisterium AI inside ChatGPT or Claude

Connect Magisterium AI via MCP to use its library as a tool in your favorite AI client.

May 11, 2026

Magisterium AI can be used inside ChatGPT, Claude, or any AI client that supports the Model Context Protocol (MCP). When connected, your AI client can use Magisterium AI's library as a source for answers.

Connect via MCP

  1. Go to magisterium.com/developers/docs/mcp.
  2. Follow the walkthrough for your client (ChatGPT custom connector, Claude MCP connection, etc.).
  3. Use the MCP server URL and an API key from your account.
The /developers page titled 'Developers' with the lead 'Leverage the knowledge of Magisterium AI in your own application! We offer various services to help you build with Magisterium AI and leverage its knowledgebase.' An 'Our Services' grid shows four cards: Chat API, Search API, MCP ('Bring Magisterium AI to your AI tools'), and A2A ('Connect AI agents to Magisterium AI'). Below it, a 'Get Started' section has API Key (Login button) and MCP Server (with a copy-pill showing 'MCP.MAGISTERIUM.COM'), plus 'Try the API' and 'Explore docs' shortcuts.
Developers page (MCP integration section)

What you get

  • Your AI client can call Magisterium AI to search the library and retrieve citations.
  • Answers inside ChatGPT / Claude can include references back to magisterial documents.
  • You keep using your AI client's interface while getting better Catholic sourcing.

Getting an API key

API keys are managed at magisterium.com/developers/api:

  1. Sign in and create a new key.
  2. Copy the key (it's shown once) and store it somewhere safe.
  3. Paste it into your AI client's MCP configuration.

Limits

  • Free: Not available — MCP and A2A both require a paid plan. Free clients receive a PLAN_REQUIRED error.
  • Pro / Organization: Unlimited requests per day, with a soft cap of 15 requests per minute (combined MCP + A2A).
  • Enterprise / Large Organization: Custom — contact us.

The full Q&A pipeline (MCP's chat tool / A2A's catholic_qa skill) is more expensive and capped at 2 requests per minute on paid plans.

The latest numbers are always on the MCP & A2A pricing page.

A2A (agent-to-agent)

We also expose an A2A endpoint for agent frameworks that don't yet support MCP — see Using Magisterium AI with other MCP / A2A clients.

Troubleshooting

  • If the connection fails, make sure your API key is correct and not revoked.
  • If responses are missing citations, your MCP client may need a specific tool to be enabled. Check the docs.

Search help

Using Magisterium AI inside ChatGPT or Claude | Magisterium AI