“Turn these three product manuals into help centers, one per product.”
- generate_upload_url
- create_file_from_temp
- app_run · pdf-to-knowledge-base
Each manual becomes its own shelf in your workspace, ready for you to review and publish.
Docsie MCP server · Knowledge base
Ask Claude, ChatGPT, Cursor or Claude Code to turn videos, PDFs or a public docs site into a knowledge base in Docsie, sort it into shelves and books, and answer your team's questions from it. The MCP server is for the people who write and run the knowledge base. Your readers use the published portal.
Runs the same Video to Knowledge Base, PDF to Help Center and Website to Help Center apps as Docsie. Uses AI credits.
https://app.docsie.io/mcpWorks in Claude, ChatGPT, Claude Code, Cursor, Codex and any MCP client.
The PDF to Help Center app in Docsie takes PDF or Word files, up to 5 at a time, and an optional name for the help center. The same app runs over MCP.
What you can ask
Your assistant runs the Docsie apps that build the knowledge base, then searches it when your team has a question. These are example requests.
“Turn these three product manuals into help centers, one per product.”
Each manual becomes its own shelf in your workspace, ready for you to review and publish.
“Build a help center from docs.example.com. Only read pages under /guides/, up to 50 pages.”
A help center built from the pages it read, named after the site unless you give it a name.
“How do we rotate our API keys? Answer from our Docsie knowledge base and name the articles you used.”
An answer drawn from the matching articles in your workspace, with each source article named so you can open it.
Real results
A screen from the Docsie app behind the second request. Over MCP, your assistant fills in the same fields and the result lands in your workspace.
The Website to Help Center app in Docsie reads a public docs site, only pages under one path if you want, up to the number of pages you choose. The same app runs over MCP.

How it works
Share a video link or upload, up to 5 PDF or Word files, or a public docs site. A site import reads 1 to 100 pages and can stay under one path.
Pick the output language and a template, then ask your assistant to create shelves and books, move articles and put them in order.
Your team asks questions in the chat and your assistant answers from the articles. Readers get the published portal, which you can style with the Publishing and access pack.
Permissions and limits
Running the apps and reorganizing content need Workspace create and edit. Portal design needs Publishing and access. Both are off by default.
Each app run uses your organization's AI credits, the same as running the app in Docsie. The MCP server itself is on every plan, Free included.
The MCP server works through your own Docsie account. A published portal has no MCP endpoint, so readers and customers use the portal itself.
See 22 example knowledge bases →FAQ
Yes. Through the Docsie MCP server they run the PDF to Help Center app on up to 5 PDF or Word files at a time. Each file becomes its own shelf in your workspace.
Yes. Website to Help Center reads a public docs site or help center, from 1 to 100 pages, and can stay under one path such as /docs/.
Yes. Video to Knowledge Base turns a recording into a searchable knowledge base. Choose a template, or a book template for a long knowledge base with its own chapter structure, and how much the writing condenses the video.
Add https://app.docsie.io/mcp as an MCP server in Cursor and sign in with your Docsie account. Cursor can then search and read the workspaces your account can already see.
No. The Docsie MCP server is for your team. A published portal has no MCP endpoint for readers, who search and read the portal itself.
With the Publishing and access pack ticked, your assistant can style and publish a portal with publish_knowledge_base. The 22 example portals on docsie.io were themed by Claude Code with that tool through a browser session, not over MCP.
Docsie MCP server
Connect Claude, ChatGPT, Cursor or Claude Code to Docsie and build, organize and search your knowledge base from the chat.