Are scanned PDFs supported?

Yes. Every PDF page is read as an image, so a scanned page is read like a page with a text layer, tables included.

How large can a PDF be?

Up to 25 MB and 500 pages a file. Split a larger file and upload the parts.

Can my AI search several PDFs at once?

Yes. search_documents searches every document in the server, or only one if your AI names it.

Does Kaiva keep my PDF files?

No. A file is held only until it has been read, then deleted. Bridge keeps the text it read and the search index.

Can the AI change my documents?

No. Document servers only read.

Which AI clients can use it?

Any MCP client, including Claude, ChatGPT, Cursor, Claude Code and VS Code. Claude and ChatGPT sign in with your Kaiva account.

GUIDE

How to turn PDFs into an MCP server your AI can search

Updated 15 October 2026

Create a documents server in Kaiva Bridge, a managed MCP cloud platform, and drop in your PDFs. Bridge reads every page, scans and tables included, and serves the PDFs as an MCP server with tools that search them and read any page in full. Add the server's address to Claude, ChatGPT, Cursor or any other MCP client; each passage your AI finds comes with its document name and page number, so it can cite them.

Start free Read the docs

Step by step

Create a documents serverSign up free at app.kaiv.ai, then choose New server, then Documents.
Add your PDFsDrop in your PDF files, up to 25 MB and 500 pages each. Word (.docx), PNG, JPEG, text and Markdown files can go in the same server. Bridge reads them in the background and shows each file's progress.
It goes live by itselfThe server publishes itself when its first document is ready. Use Try a search in the console to see the passages your AI will get back.
Connect your AI clientCopy the address from the Connect tab and add it to Claude, ChatGPT, Cursor or another MCP client. Claude and ChatGPT sign in with your Kaiva account, so no key is needed.

From a terminal, for one PDF, several, or the files directly inside a folder. It waits until every file has been read:

Terminal

npx @kaiva/bridge login
npx @kaiva/bridge push product-manuals --docs ./manuals
npx @kaiva/bridge install product-manuals --client cursor

How Bridge reads a PDF

Each page is turned into an image and read by a document-parsing model, so a scanned page is read the same way as a page with a text layer, and tables come back as tables (as Markdown tables, in what your AI receives). The text is then split into passages and indexed for search.

Bridge keeps the text it read and its search index, not your PDF: the file is held only until it has been read, then deleted.

The tools your AI gets

search_documents finds the passages that answer a question, five by default and up to ten, each with its document name and page number. It searches every PDF in the server, or only one if your AI names it.

read_document reads one page in full when a passage is not enough, and list_documents lists the documents that are ready, with their page counts.

You can also save a search as a tool of its own, such as warranty_terms, optionally pinned to one document, so your AI calls it by name. A server can have up to 20 saved searches.

Then ask your AI client something like: "What do the manuals say about cleaning the filter? Quote the page." It searches, reads the page if it needs more, and answers with the document and page.

Updating and removing PDFs

Upload a file with the same name to replace that document. The old version stays searchable until the new one is ready, and an update you no longer want can be cancelled. Delete a document to remove it from the server and its search.

How pages are counted

Every page of a PDF counts toward your plan's page allowance: 200 pages on Free and 20,000 on Pro. An image counts as one page, and Word, text and Markdown files are counted in pages of about 3,000 characters.

If a PDF is refused

Password-protected. Remove the password and upload the file again.

Over 25 MB or 500 pages. Split the file and upload the parts.

Damaged, or larger inside than Bridge reads. Save it again from the app that made it, or split it.

Over your page allowance. Delete documents you no longer need, or upgrade to Pro.

What you need

Your PDFs, up to 25 MB and 500 pages each. No database, vector store or server of your own is needed.

Free includes 10 MCP servers, 200,000 tool calls a month and 200 pages of documents, with no card.

Other ways to do it

Attach the PDF in one appClaude, ChatGPT and other AI apps take PDFs in a chat or a project. That works inside that app; a documents server is one address that every MCP client can search.
Build your own retrieval pipelineParse the PDFs, split and embed the text, store it in a vector database and write an MCP server around the search, with open-source tools for each step. Full control, and a pipeline to run.

Questions

Are scanned PDFs supported?: Yes. Every PDF page is read as an image, so a scanned page is read like a page with a text layer, tables included.
How large can a PDF be?: Up to 25 MB and 500 pages a file. Split a larger file and upload the parts.
Can my AI search several PDFs at once?: Yes. search_documents searches every document in the server, or only one if your AI names it.
Does Kaiva keep my PDF files?: No. A file is held only until it has been read, then deleted. Bridge keeps the text it read and the search index.
Can the AI change my documents?: No. Document servers only read.
Which AI clients can use it?: Any MCP client, including Claude, ChatGPT, Cursor, Claude Code and VS Code. Claude and ChatGPT sign in with your Kaiva account.
How much is free?: Free includes 10 MCP servers, 200,000 tool calls a month and 200 pages of documents, with no card. Pro includes 20,000 pages.