The ClickHelp Authoring MCP Server is a service that allows AI agents and LLM-powered tools — such as Claude, Gemini, ChatGPT, and others — to interact with your documentation portal programmatically. It implements the Model Context Protocol (MCP), an open standard for connecting AI assistants to external data sources and tools.
Think of it as an API layer designed specifically for AI agents: instead of parsing HTML or guessing your portal's structure, an agent can query the Authoring MCP Server directly to read topics, search content, and even make changes.
Licensing
The Authoring MCP Server is a part of the REST API & MCP add-on.
How It Works
The Authoring MCP Server acts as a bridge between an AI agent and your ClickHelp portal. When an agent needs to retrieve or update content, it sends requests to the Authoring MCP Server, which communicates with the portal through the REST API and returns the result to the agent.
A typical interaction looks like this:
- An administrator configures the AI agent to use the ClickHelp Authoring MCP Server.
- A user sends a request to the agent in natural language — no special commands required.
- The agent decides whether to use the Authoring MCP Server to fulfill the request and which tools to invoke.
- If this is the first time the agent uses the server, a browser tab opens with the standard ClickHelp login page. The user authenticates using their regular credentials — SSO and 2FA are fully supported. No API keys are required.
- After login, the agent stores an authentication token and uses it for subsequent requests. Users are asked to re-authenticate every 30 days.
-
The agent acts on behalf of the authenticated user, with that user's exact permissions.
We recommend creating separate Contributor or Power Reader accounts for your AI agent so that you can track the AI-made changes and set specific permissions. See Setting Up a Dedicated AI Account.
The Authoring MCP Server endpoint for a portal looks something like this:
https://us.mcp.clickhelp.com/portals/yourportal.clickhelp.co,
and can be obtained in your portal settings, AI integrations → MCP server.
To see what you can accomplish with an AI agent and the Authoring MCP Server, refer to Authoring MCP Server Use Cases.
Setting Up a Dedicated AI Account
To use a dedicated account for the AI agent — which is recommended for audit and permission control — follow these steps:
- Create a separate Contributor or Power Reader account for the AI agent.
- Log out of your portal in the current browser and log back in with the dedicated AI account — or open a separate browser profile/incognito session and log in there.
- Connect the Authoring MCP Server to your AI tool and make a first request to access your ClickHelp portal.
- After the connection is established, the agent stores a token for this account and uses it for the next 30 days.
| The portal session is only used once — at the moment of the initial connection. After that, the agent works with a stored token and does not depend on which account is currently logged in to the portal in your browser. |
Security
Access through the Authoring MCP Server is governed by the same permission system as direct portal access. An agent can only perform actions that the authenticated user is allowed to perform. This means you can create a dedicated portal account with limited permissions for automated workflows, restricting what the agent can read or modify.
Authentication between the agent and the Authoring MCP Server uses OAuth 2.1. The connection between the Authoring MCP Server and your portal is secured by ClickHelp. Note that data passes from the portal through the Authoring MCP Server to the LLM you are using — the choice of LLM and its data handling are the responsibility of your organization.
Available Tools
The Authoring MCP Server provides two categories of tools.
Read-only tools allow agents to:
- List projects, publications, and translation projects in the portal
- List topics within a project or publication, including API documentation topics
- Retrieve topic content
- Retrieve the table of contents structure
- Search portal content
- Query AnswerGenius
- Retrieve topic statuses and available workflow transitions
Write tools allow agents to:
- Create projects
- Create and edit topics, including workflow fields (status, assignee, owner)
- Modify the table of contents structure
The following actions are not currently supported: deleting topics or projects, importing content using the ClickHelp import engine, publishing content, changing project visibility, and modifying portal settings, home pages, styles, or scripts.
| While the built-in import engine is not accessible via the Authoring MCP Server, you can ask an agent to process content in any format it can parse — including PDF, Excel, or PowerPoint files — and create topics from that content directly. |
Topics in API documentation projects cannot be created or edited, and only workflow fields can be modified in translation projects.
Rate Limits
There are the following limits for the Authoring MCP server tool requests:
|
Limit
|
Value
|
Comments |
|---|---|---|
| MCP tool calls per portal per day | 20 000 | MCP tool calls from all users of the same portal per day. |
|
MCP tool calls per portal per minute
|
180 |
MCP tool calls from all users of the same portal per minute. |
|
MCP tool calls per IP address per minute |
60 | MCP tool calls per user (IP address) per minute. |
| Read tool requests per second | 10 | Requests when an MCP read-only tool reads information. |
| Write tool requests per second | 10 | Requests when an MCP write tool updates content. |
| Long tasks concurrent count | 3 | Long tasks - when an MCP tool performs an action with a project or publication. Currently, it can be only creation of a project. |
- MCP tool call - one successful request of a read or write tool (for example, searching for a topic or creating one).
- Limits for read and write MCP tool requests are tracked separately.
Limit Is Reached
When a limit is reached, the AI agent is told that it has made too many requests and should wait before trying again. Most AI agents handle this on their own — they pause for a moment and retry automatically — so in most cases you will not see anything go wrong.
If you do see a rate limit error in your AI agent, here is what to do:
- Wait about a minute, then try again. The per-minute limits reset every minute, so a short pause is usually enough.
- Check whether teammates are working on the same portal at the same time. Some limits apply to the whole portal, not just to you. If several people are heavily using AI agents on the same portal at once, their combined activity can reach the portal limit.
If your portal regularly needs more capacity, higher limits are available through the REST API & MCP: High Volume add-on. See the Pricing page for details.