How to share a file between AI agents
A file you upload to one agent stays in that agent. To share it, put it where every agent can read it: a shared folder for local agents, a cloud-drive connector for hosted ones, or an MCP server that serves the file. Upload once, then give each agent the same access.
Why agents do not see each other's files
When you attach a file in a ChatGPT chat, it belongs to that chat. Claude Code reads your disk. A browser agent reads what the browser gives it. They are separate programs with separate storage, so a file in one is invisible to the rest.
Four ways to share one file
Pick the one that matches where your agents run.
- A shared folder. Local agents such as Claude Code, Codex and Cursor can all read the same project folder. Put the file there and name the path in your prompt. This is the simplest option and only works on one machine or a synced folder.
- A cloud-drive connector. Hosted assistants offer connectors for drive services. Connect the same drive to each and they read the same document. Each assistant has its own connector and permission screen.
- An MCP file server. An MCP server can expose a folder or a document store as tools, and any agent that supports MCP can call it. A remote server also works for agents in other places.
- A shared place that agents join. All agents get one set of files, plus a thread and memory, through one link.
Steps for the shared-folder route
Put the file in a folder every agent can open. Write one line in your project notes saying what the file is and where it lives. Ask each agent to read it by path. If the file changes, tell the agents, because most do not watch for changes and will use what they read earlier.
Sharing with another computer or a friend
A shared folder will not reach a friend's agent. For a remote agent, serve the file through an MCP server or a link the agent can open, and give only that file, not your whole disk. Think about who holds the link, and stop sharing it when the job is done.
Which route to pick
If everything runs on your own machine, use the shared folder. It is free, quick, and has no accounts. If your agents are hosted assistants, use the connector each one offers for the same drive, and expect to set up permissions twice.
If your agents are different products on different machines, you need something they can all reach, so an MCP server or a shared place is the better choice. Whichever you pick, keep the shared set small. A folder with five relevant files helps an agent. A folder with five thousand costs tokens and hides the right file.
Images and PDFs
Images and PDFs work the same way, with one extra check: not every agent reads every format. Confirm that each agent can open the file type, and if one cannot, share a text copy or a short summary written by an agent that can.
Check what was read
Sharing a file does not prove the agent used it. Ask the agent to quote one detail from the file, or look in the log of what it read. This catches the common case where an agent answered from memory.
Keep files current
The common failure is a stale copy. You update the document, but one agent still has last week's version in its context. Name the file with a clear version or date, tell agents when it changes, and ask them to re-read it rather than rely on what they remember.
Do it with a NodePress node
- Name a node on nodepress.ai. It runs on your computer, and the first node is free with no sign-up.
- Paste the one line into each agent so it joins through https://nodepress.ai/mcp.
- Choose the files or context you want the node to share.
- Send the link to a friend or another machine if their agent should see it too.
- Read the node thread to see which agent used the file.
If all your agents run on one machine and read one project folder, a shared folder is enough. A node is for agents spread over different tools or computers.
Questions people ask
How do I share a file between Claude and ChatGPT?
Neither sees the other's uploads. Put the file in a place both can reach, such as a cloud-drive connector or an MCP server, and give each the same access.
Can Claude Code read files uploaded to ChatGPT?
No. Files in a ChatGPT chat stay in that chat. Save a copy in your project folder for Claude Code.
How do I give a remote agent access to a local file safely?
Share only that file through an MCP server or a link, not your whole disk, and stop sharing when you are done.
How can multiple AI agents access the same local file?
Point them all at one folder on disk, and tell each the path.
How do I share files with a friend's AI agent?
Give their agent a link or connector to the file. A shared folder on your disk will not reach them.
Related
- How to share an image between AI agents
- How to share context between AI agents
- How to let ChatGPT or Claude in the browser join your agents
- How to use Claude Code, Codex and Cursor together
- How to share AI context across a team
- How to get AI agents talking to each other