This article covers how to connect your AI client to the Databook MCP server, what to expect during sign-in, and how to check the connection is working. For what the server can actually do once connected, see Databook MCP server.
Connecting requires an active Databook account with an MCP entitlement, and credentials issued by Databook. If sign-in fails with a permissions error, check your entitlement first.
Before you start
You'll need:
An AI client that supports the Model Context Protocol over streamable HTTP, for example the Claude desktop app or Claude Code
A Databook user account, on a tenant with MCP access enabled
Permission to add connectors in your AI client. In managed deployments this is often restricted to an administrator, who may need to approve the connector for your organization before individual users can enable it
Configuration values
Setting | Value |
Server URL | |
Transport type | Streamable HTTP |
Authentication | OAuth |
Connection | Direct |
No API key, secret, or token needs to be pasted into the client. Authentication happens through an OAuth sign-in in your browser, and the client stores the resulting credential itself.
Setting it up in the Claude desktop app
Open Settings, then Connectors.
Choose Add custom connector.
Give the connector a name, for example Databook.
Enter the server URL: https://tools.databook.com/mcp.
Save the connector. Claude will show it in the connector list as not yet connected.
Choose Connect. A browser window opens for Databook sign-in.
Sign in with your Databook account and approve the access request. The browser returns you to the client, and the connector shows as connected.
Setting it up in Claude Code
Add the server, specifying HTTP transport: claude mcp add --transport http databook https://tools.databook.com/mcp
Start Claude Code and run /mcp.
Select the Databook server and choose to authenticate. A browser window opens for Databook sign-in.
Sign in and approve the access request, then return to the terminal. Running /mcp again shows the server as connected.
Other AI clients
Any client that supports MCP over streamable HTTP with OAuth can connect using the same values. The wording differs between clients: look for an option to add a custom or remote MCP server, then supply the server URL and let the client handle the OAuth exchange. Clients that only support local MCP servers launched as a subprocess aren't compatible, since the Databook server is remote.
Checking the connection works
Confirm your client lists the Databook tools. Most clients show the available tools alongside the connector once it's connected.
Ask a question that needs Databook, for example "Using Databook, summarize the latest earnings call for Nike." Naming Databook in your request makes the client more likely to route to the right tool on the first try.
Check that the answer includes Databook content and citations, rather than the client's own general knowledge.
Some tools run asynchronously, so your client may take longer on those requests while it waits for the job result. This is expected behavior, not a failed connection.
Troubleshooting
Symptom | What to check |
Connector saves but never connects | Confirm the URL is exactly https://tools.databook.com/mcp and that the client is set to HTTP transport rather than a local command |
Sign-in completes but access is denied | Your Databook account is likely on a tenant without MCP access, or without the entitlement. Contact your Databook account team |
Browser sign-in window doesn't open | Check for a blocked pop-up, and that the client has permission to open a browser. Retrying the connect action usually resolves it |
Tools are listed but never used | Name Databook explicitly in your request. Some clients need the tool to be enabled for the specific conversation as well as installed |
Connection worked previously and now fails | Your stored credential may have expired. Disconnect and reconnect the connector to run sign-in again |
Answers are slow or time out | Asynchronous tools take longer by design. If it persists across tools, check whether your request is unusually broad and try a more specific question |
Getting help
Setup instructions maintained alongside the Databook plugin, including the latest client-specific steps, are published in the Databook plugin repository. For entitlement and credential questions, contact your Databook account team.