GitBook Assistant

GitBook Assistant offers users answers based on your docs and tailored to their situation — not just generic responses

This feature is available on the Ultimate site plan.

The GitBook Assistant

GitBook Assistant gives your users fast, accurate answers about your documentation using natural language. It’s personalized on your users, can be embedded into your website or product, and available in the sidebar of your docs.

Think of it as a product expert available to all of your users, in the places and times they need it most.

The Assistant uses agentic retrieval to understand the context of the query based on the user’s current page, previously-read pages, and previous conversations they’ve had.

GitBook Assistant is trained on your documentation, but you can also add external sources to expand it’s context and knowledge and give better answers.

Test GitBook Assistant

Using GitBook Assistant in published docs

• Press ⌘ + I on Mac or Ctrl + I on PC

• Click the GitBook Assistant button next to the Ask or search… bar

• Type a question into the Ask or search… bar and choose the ‘Ask…’ option at the top of the menu.

Extend GitBook Assistant with MCP servers

You can also choose to add external data sources to GitBook Assistant to give it more context and data to pull answers from. You can do this by connecting Assistant to MCP servers for external platforms, such as:

• Your community (Slack, Discord, GitHub Communities etc)

• Support tools (Intercom etc)

• Your future product roadmap (GitHub, Linear etc)

• Docs for external integrations with products

To add an MCP server to GitBook Assistant, follow these steps:

1

Open your site’s settings

Navigate to your site dashboard and choose the Settings option from the site header. Then choose the AI & MCP section from the left-hand menu

2

Add a new server

At the bottom of the page is a table showing all the connected MCP servers. To add a new server, click Add MCP server

3

Choose your MCP server

To add your server you’ll need to give it a name, add the URL for the server, and configure the HTTP headers that will be sent along with the request to the server when a user submits a query.