Model Context Protocol

The knowledge base MCP server for your AI assistant.

HelpDocs MCP connects Claude, ChatGPT, and any other MCP-compatible client straight to your knowledge base software, so your assistant can search it, read it in full, and draft into it.

Searched your articles for “SSO setup”Claude
Read “Refund policy” in fullChatGPT
Returned 12 matching articlesHelpDocs MCP
Drafted “Webhook retries” from a changelogClaude
Capabilities

What your assistant can do once it's connected.

An assistant is only as good as what it can reach. MCP gives it the real knowledge base, live, rather than whatever it managed to crawl.

Search your knowledge base

Ask a question in plain language and the assistant searches your articles and categories, then works from what it finds instead of guessing.

Read the whole article

Not a snippet, not a scrape. The full, current content of any article, so answers match what your readers actually see.

Draft and update articles

Create, update, and delete articles and categories, drafts included, so a rewrite is a conversation rather than an afternoon.

Check what's working

Pull analytics and account details without leaving the assistant, so you can ask which articles get read and which ones quietly fail.

Work in any MCP client

MCP is an open standard, so one connection covers Claude, ChatGPT, and whatever your team adopts next. No per-tool plugin to maintain.

Stay inside your permissions

The connection is authorized against your HelpDocs account, and you choose which account to connect if you run more than one.

Any MCP client

Connect it once, use it everywhere.

MCP is an open standard, so HelpDocs does not need a bespoke plugin for every assistant on the market. Set the connection up in the client you already use, and the same server answers when your team adopts a different one next quarter.

See the MCP integration

Connected in about five minutes

  1. 1Open your AI client and add a new MCP connection
  2. 2Name it HelpDocs and paste your HelpDocs MCP URL
  3. 3Follow the prompts to authorize MCP usage
  4. 4Pick which HelpDocs account to connect if you have more than one
  5. 5Ask your assistant to search your HelpDocs articles to confirm it works
Read the full setup guide
Prompts to try

Eight things to ask on day one.

Connect HelpDocs first, then open one of these in your assistant. Each one leans on a different part of the server, so together they are a decent tour of what it can reach.

Audit
Search my HelpDocs knowledge base for articles about refunds and tell me where any two of them contradict each other.
Find the gaps
Look at my HelpDocs categories and list five articles a brand new customer would go looking for and not find.
Answer a ticket
A customer asks how to move their account to a new domain. Answer using only my HelpDocs articles, and cite the ones you used.
Rewrite
Read my HelpDocs getting started article, then draft a shorter version for someone who has never used the product before.
Catch stale wording
Find every HelpDocs article that still refers to a feature by its old name, and give me the exact wording to change.
Ask the analytics
Pull my HelpDocs analytics and tell me which five articles get the most views but have the least content.
Draft from a changelog
Turn these release notes into a draft HelpDocs article that matches the tone of my existing docs, and leave it unpublished.
Rethink the structure
Review my HelpDocs category structure and suggest a layout that matches how customers actually phrase their questions.

Want a longer prompt?

Every one of our free document templates ships with a full generation prompt, ready to open in Claude or ChatGPT. Point it at your connected knowledge base and the draft lands in HelpDocs.

Browse templates
The wider ecosystem

Your docs, in the same room as everything else.

An assistant can hold several MCP connections open at once. Add HelpDocs next to the thousands of other servers people run for issue trackers, repos, design tools, databases, and analytics, and it can answer using your documentation and your systems in the same breath.

Write from what shipped

Pair HelpDocs with your issue tracker and ask for a draft of the release note that the changelog implies.

Close the gaps

Pair it with your support inbox and ask which repeat questions have no article behind them yet.

Keep it current

Pair it with analytics and ask which popular articles have not been touched since the last redesign.

A connection is only as good as what it connects to.

MCP does not make documentation true. It makes whatever you have reachable, at machine speed, with a confident voice on top. A stale article wired into an assistant becomes a stale answer given to every customer who asks.

That is the part HelpDocs spends most of its time on: keeping articles current, keeping the wrong people out of the private ones, and showing you which pages are quietly failing.

FAQ

MCP questions, answered

What is a knowledge base MCP server?
Model Context Protocol (MCP) is an open standard for connecting AI assistants to outside systems. A knowledge base MCP server exposes your documentation as tools the assistant can call, so it can search articles, read them in full, and write new ones. HelpDocs MCP is that server for your HelpDocs knowledge base.
Which AI assistants work with HelpDocs MCP?
Any MCP-compatible client. Claude and ChatGPT are the two most people start with, and because MCP is an open standard the same connection works in other clients that support it. There is nothing to rebuild when you switch tools.
Can my assistant edit articles, or only read them?
Both. The server exposes search and read alongside create, update, and delete for articles and categories, including drafts. A common pattern is to let the assistant draft into an unpublished article so a human still reviews it before it goes live.
How is this different from pointing a chatbot at my help center URL?
Pointing a bot at your public help center gives it whatever it can crawl, and only what is public. MCP gives the assistant live, structured access to the knowledge base itself, so it reads current content, sees categories, and can write back. If a crawl is all you need, HelpDocs also offers a chatbot source page.
Do I need a developer to set it up?
No. You add a new MCP connection in your AI client, paste your HelpDocs MCP URL, and authorize it. The setup guide walks through it, and the whole thing takes a few minutes.

Give your AI assistant something worth reading.

Start free 14-day trial

Connect HelpDocs over MCP and let Claude, ChatGPT, or any other client work from documentation you actually maintain.

No credit card required.

HelpDocs onboarding example