How to Give ChatGPT Access to Local Files

ChatGPT can read local files through two main routes: file uploads in the web or desktop app, and a desktop app integration that gives it context from what is open on your screen. Tools like fenbs keep the project goals and source file references that AI apps should carry from session to session, reducing how often you need to re-upload files and re-explain what the project is about.

This guide covers both approaches, what ChatGPT can do with local files, what it cannot do regardless of how you connect it, and how to avoid the re-upload problem for ongoing projects.

File uploads: the most reliable method

The simplest way to give ChatGPT access to a local file is to upload it. In ChatGPT on the web or in the desktop app, use the attachment button or drag the file into the conversation. ChatGPT can read PDFs, Word documents, Excel spreadsheets, PowerPoint slides, text files, code files, and images.

Uploads are scoped to the current conversation. When you start a new chat, ChatGPT does not have that file. If you need it again, upload it again. This is the right method for most one-off tasks: reviewing a contract, analyzing a dataset, or explaining a piece of code. It is also the safest approach because you control exactly what ChatGPT sees.

For large files, check that ChatGPT processed the full document rather than just the opening section. Ask it to quote from a part of the file you know is toward the end. If it cannot, the file may have been truncated.

The desktop app and screen context

The ChatGPT desktop app can read content from other apps on your screen in some configurations. When you ask a question about something you have open, the desktop app may be able to see the active document, selected text, or the name of the app you are using.

This is different from giving ChatGPT access to a folder on your drive. It does not browse your filesystem. It reads structured context from what is currently visible. Check OpenAI's current documentation for what the desktop app can and cannot access, since this changes over time.

What ChatGPT cannot do with local files

ChatGPT cannot browse your hard drive, read files you have not explicitly shared, or write files back to your local filesystem through a standard connection. It cannot monitor a folder for changes or pull in an updated version of a file automatically. If you update a file and need ChatGPT to see the revised version, upload it again in the current or a new conversation. There is no sync mechanism between your local drive and ChatGPT; each upload is a one-time transfer for that session.

For accessing files stored in Google Drive, see how to give ChatGPT access to Google Drive. For GitHub repositories, see how to give ChatGPT access to a GitHub repo. For comparison with Claude's approach, see can Claude Desktop access local files.

Choosing between upload and connected storage

File uploads make sense when you need to analyze a document once and move on. Connected storage, such as Google Drive, makes more sense when you need to reference the same files regularly across multiple sessions or share them with a team. Evaluate by how often you return to the same files and whether others need the same access.

Avoid uploading the same file repeatedly across sessions as a workaround for ChatGPT's lack of memory. That pattern works for a while but adds friction and does not help with decisions made about the file in earlier sessions. A structured project record is a more durable solution.

Keeping your project context across sessions

The main limitation of file uploads is that they reset with each conversation. If you are working on a project across multiple sessions, you either re-upload the files each time or find a way to keep your project context available without relying on disk access.

Saving the project goal, the key file references, and the decisions made about those files in fenbs gives each session a starting point. A connected assistant reads the project context and knows which files to ask for, which version is current, and what was already decided. You upload only the specific file needed for the current task.

How to do this with fenbs

1. Create a free account. Your personal board is free and needs no credit card.
2. Add a project for the work that uses those local files. Write the goal and list the key files by name or location.
3. Record decisions about those files in Decisions and rules: which version is current, which sections are approved, what format is required.
4. Connect ChatGPT to fenbs over MCP from Integrations. ChatGPT connects as a developer app; check which plans include MCP support.
5. At the start of each session, ask ChatGPT to read the project context. Upload only the specific file it needs for the current task.

fenbs does not access your local filesystem. It keeps the structured context that reduces re-uploading and carries the decisions from one session to the next.