Skip to main content
Table of Contents

Connect HelpDocs to Claude

Connect Claude to HelpDocs via MCP to read and manage Knowledge Base content. Three-step setup, configure permissions for read/write actions, control what Claude can do automatically or with your approval.

River Sloane
Updated by River Sloane
Table of Contents

Model Context Protocol (MCP) lets you connect Claude to your HelpDocs account so it can read, search, and work with your Knowledge Base content directly. This guide shows you how to set up the connection with Claude and what you can do with it.

What is MCP? 🤔

MCP (Model Context Protocol) is a way for AI assistants like ChatGPT to connect to external tools and data sources. It gives ChatGPT the ability to interact with your HelpDocs content in real-time. Learn more about it in our article over here.

Connecting Claude to HelpDocs MCP

Set up in Claude web or Claude Desktop
  1. Click here to head to the HelpDocs connector listing or search for HelpDocs in the directory
  2. Click Connect for your team
HelpDocs logo with Community and New badges, captioned 'The knowledge base Claude can read and write'
Authorize use with HelpDocs

Follow the authorization flow. This connector uses OAuth to read and write to your knowledge base.

If you have multiple Knowledge Bases with us, you'll need to select one to connect to.
Start using HelpDocs with Claude

Once you've added the connector, Claude will be able to access your Knowledge Base content whenever you're chatting with it. Just ask to use HelpDocs 🙌

Setting Up Tool Permissions

You can control exactly what Claude is allowed to do with your HelpDocs content. This lets you decide which actions happen automatically, which need your approval first, and which aren't allowed at all.

To configure permissions:

  1. Click on the HelpDocs connector in Claude
  2. Head to Tool permissions
  3. Set your preferences for each action
Screenshot of Claude tool permissions for MCP

You can choose from three permission levels:

  • Automatic: Claude can perform the action without asking
  • Manual approval: Claude will ask for your permission before doing it
  • Not allowed: Claude can't perform this action at all
We recommend setting write actions (like creating or editing articles) to manual approval when you're first getting started. This gives you control while you learn how Claude interacts with your Knowledge Base.

Actions Available to Claude

Here's everything the connector can do, grouped by what you'd use it for. Read actions only look at your content. Write actions change it, so those are the ones worth setting to manual approval.

What you want to do

Actions

Read or write

Search and read your content

search, fetch, list_articles, list_categories, get_category

Read

See how your content is performing

get_analytics

Read

Check who has access to what

list_permission_groups

Read

Draft something new

create_article, create_category

Write

Edit content you already have

update_article, update_category

Write

Push edits live

publish_working_copy

Write

Remove content

delete_article, delete_category

Write

Recover something deleted

list_deleted_articles, restore_article

Read and write

Editing a published article doesn't change your live Knowledge Base straight away. Claude stacks the edits up as a working copy, and your help center stays exactly as it is until the changes are published.

What did you think of this doc?

Connect HelpDocs to ChatGPT

Using an API Key with HelpDocs MCP

Get in touch