Connect an AI assistant over MCP

Any assistant that can use a remote MCP server can read CORA's governed context, including Claude, ChatGPT, Microsoft Copilot, and your own agents. The assistant is your company's choice; CORA's side is the same for all of them.

What the assistant needs

• The endpoint. CORA's MCP server speaks Streamable HTTP. For the hosted product the address is https://hellocora.eu/api/mcp. The MCP access page shows it under Connection -> Streamable HTTP endpoint.
• An access token, sent as a bearer token in the request's Authorization header: the word Bearer, a space, then the token.

CORA does not offer a sign-in (OAuth) flow for MCP today. The assistant must let you set that header for a custom MCP server. If an assistant only accepts connectors that sign in through OAuth, it cannot connect with a CORA token yet.

On the CORA side

1. A workspace owner or admin opens Settings -> MCP access.
2. Under Create access token, give the token a Name that says which assistant uses it, set Expires in days, and leave on only the Allowed tools that assistant needs.
3. Choose Create token and copy it with the copy button. It will not be shown again.

On the assistant's side

In your assistant's settings, add a custom MCP server (some products call it a connector or an action). Paste the endpoint, choose Streamable HTTP if you are asked for a transport, and add the Authorization header with your token. The exact menus differ between products and change over time, so follow your assistant's own documentation.

Check that it works

Ask the assistant something your company memory can answer. In CORA, the call appears under Recent calls on the MCP access page, with the token name, the tool used, and its status. A call with an invalid, expired, or revoked token is refused.

Use one token per assistant, so you can see each one's calls and revoke one without breaking the others.