ChatGPT Connectors for OneDrive
You can get OneDrive files into ChatGPT through file uploads or a custom connector using the Microsoft Graph API. Here is what each route allows, the permission checks to run, and how to keep your reviewed source links.
Updated 5 min read
You can get OneDrive files into a ChatGPT session by uploading them to a conversation or by building a custom connector using the Microsoft Graph API. There is no first-party OneDrive connector in ChatGPT's standard catalog. Understanding what each route gives ChatGPT access to helps you avoid a setup that looks connected but cannot see the files you need. Tools like fenbs can keep the reviewed source links and project decisions in a place that lasts beyond any single chat session.
This page covers the two main options, the access each provides, and the permission checks worth running before you rely on either.
Upload OneDrive files to a conversation
The simplest route is to download a file from OneDrive and upload it to a ChatGPT conversation. ChatGPT can read Word documents, PDFs, plain text files and spreadsheets you attach. You can ask questions about the content, extract key points, or use it to draft a response.
This works without any connector setup and on any plan that supports file uploads. The limit is that it is a manual transfer: ChatGPT has no ongoing access to OneDrive, and if the file changes after you upload it, your copy is out of date.
Build a custom connector with the Microsoft Graph API
Microsoft Graph is the API layer for OneDrive (and most Microsoft 365 services). It uses Microsoft Entra (formerly Azure Active Directory) for OAuth 2.0 authentication. You can build a custom connector on the ChatGPT connector platform that authenticates with Graph and lets ChatGPT search OneDrive, list folders, and retrieve file content during a conversation.
Building a connector requires developer mode in ChatGPT. Which plans include it is OpenAI's to confirm: check OpenAI's help center (opens in a new tab). For the underlying connector mechanics, including OAuth registration, see chatgpt connectors api.
Permission checks to run before connecting
Microsoft Graph has a detailed permission model. Run these checks before building or testing a connector.
- API permissions: register an app in Microsoft Entra and request the Files.Read or Files.ReadWrite permission scopes for OneDrive access. An app with no file permissions will return an authorization error.
- Admin consent: for work or school accounts, some Graph permissions require admin consent. A personal (consumer) Microsoft account does not need admin consent. Work accounts do, for most permissions.
- Token expiry: Microsoft access tokens expire after one hour. Your connector must refresh them using the refresh token.
- Shared files: the connector can only read files the authenticated account has access to. Files shared with the account appear in the Shared section of OneDrive and are accessible, but files owned by others and not shared are not.
- SharePoint overlap: OneDrive for Business stores files in SharePoint under the hood. Files created in SharePoint team sites are not in a user's personal OneDrive and need different Graph endpoints.
Keeping reviewed source links and project decisions
A ChatGPT session that reviews OneDrive documents often produces conclusions: this version is current, that clause needs a second opinion, the numbers in the spreadsheet are approved. Those conclusions belong outside the chat, not in a transcript that disappears next session.
fenbs keeps those conclusions as Links and listings (the specific OneDrive documents reviewed with their status), Decisions and rules (what was accepted and what was flagged), and task progress (what still needs review). For privacy considerations when connecting Microsoft storage, see chatgpt connectors privacy.
OneDrive is widely used for sharing documents between team members, which means a document review workflow often involves more than one person at different times. Without a shared record of which version was reviewed and what was found, two reviewers may examine the same document independently and reach different conclusions, or a later reviewer may approve something that an earlier one flagged. The review record in fenbs removes that ambiguity. Each document link carries its review status and the key finding, so the next person who opens the project can see at a glance which documents are cleared, which are flagged, and which have not been reviewed yet. They do not need to contact the person who ran the earlier sessions to find out what was already done. For Microsoft 365 environments where OneDrive is the primary document store, this record also provides a lightweight layer of governance over AI-assisted review, without requiring a separate review management system.
Track which OneDrive documents were reviewed and what was found
Add each reviewed OneDrive document as a link in fenbs with its review status and key finding. Write the accepted decisions as Decisions and rules. Connect ChatGPT from Integrations and ask it to read the fenbs context before each session, so it knows which files were already reviewed and what was concluded.
Create a free accountFree for one person · No credit card · Works with Claude, ChatGPT, Cursor and more
How to do this with fenbs
- Create a free account. One person is free with no credit card.
- Add your project and the documents under review as tasks or links.
- For each reviewed document, record the finding and outcome as a Decisions and rules entry or a task note.
- Connect ChatGPT to fenbs from Integrations using the MCP setup guide.
- At the start of each session ask ChatGPT to read the fenbs context. Before stopping, ask it to record any new findings and update the review status of documents it worked on.
fenbs does not connect to OneDrive or SharePoint. It keeps the document links, findings, and project decisions your AI sessions build on. For a comparison of ChatGPT connector options across tools, see chatgpt connectors and apps and chatgpt connectors not working.