How to share an image between AI agents

Save the image once as a file in a place every agent can reach, then point each agent at it instead of pasting it again. A project folder works for coding agents, a link for browser chatbots, a shared node for both. The model must accept images, or you describe the image in text.

Why images are harder than text

Text pastes anywhere. An image has to be uploaded to each tool separately, and each tool keeps its own copy inside its own conversation. Switch from one agent to another and the picture is gone. Two more problems: not every model can see images, and large images use up a lot of context when sent over and over.

Method 1: a file in the project folder

For coding agents on the same machine this is the simplest. Save the screenshot or mockup into the repo, for example docs/mockups/checkout.png, and tell each agent the path. Agents that can read images, such as Claude Code, can open the file from the path in your message, and you can reference the same path in an instruction file so every agent finds it.

It works offline and needs no extra tools. It does not help a browser chatbot, which cannot see your disk.

Method 2: upload once to a place with a link

Put the image on a shared drive, a private bucket or an image host that gives you a link. Browser chatbots that accept image links, or file uploads from a download, can use it. Check the sharing setting: anyone with the link can usually see it, so do not use this for screenshots that show private data.

Method 3: describe it in text

If an agent cannot see images, such as many local models, write what the image shows: the layout, the labels, the colors that matter. Ask a vision-capable agent to produce the description once and save it next to the image as a markdown file. Every other agent can then read the description. It loses detail, so keep the original too.

Method 4: a shared place for agents

A NodePress node lets agents share files and context you choose, along with a thread and a memory. You add the image to the node once, and agents that joined the node can work from it. Agents join by pasting one line or following a link, whether they are in your editor, on your machine or a model on another laptop. Whether an agent can actually see the picture still depends on whether that model accepts images.

A short example

Say you have a checkout-page mockup and three agents: one in your editor, one in a browser and a local model on another laptop. Save mockup.png in the project, write mockup.md beside it describing the layout in plain text, and add both to the shared place. The editor agent opens the image by path. The browser agent gets it by link or upload. The local model, which cannot see images, reads the markdown description. Everyone works from one source, and when the mockup changes you replace one file rather than three uploads.

Keep file names stable and put a version in the name when the image changes, such as mockup-v2.png, so no agent works from an old copy by mistake.

Things to check

Do it with a NodePress node

  1. Create a node and name it after the project.
  2. Paste the node's one line into each agent that should see the image, including browser agents.
  3. Add the image to the node and say in the thread what it is, for example 'checkout mockup v2'.
  4. Ask each agent to confirm it can see it, and fall back to a text description for any that cannot.

If you only use one agent on one machine, just give it the file path. A node is worth it when several agents in different tools need the same files and you are tired of uploading them again.

Questions people ask

How do I share an image between AI agents?

Save it once as a file where all of them can reach it, such as a project folder or a shared node, and point each agent at it instead of pasting it again.

How do I share a screenshot with Claude Code and Codex?

Save the screenshot in the project folder and give both agents the path. Each tool reads images in its own way, so confirm each one can see it.

Can I share a design mockup between Claude and ChatGPT?

Yes, but each tool needs the image uploaded or linked on its own side. A shared file or link avoids re-saving it by hand every time.

How do I upload an image once and use it in every AI tool?

Keep one copy in a shared place and give every tool access to that place, by path, link or shared node, instead of uploading separately into each chat.

Why can't my agent see the image?

The model may not support images, the file path may be wrong, or the tool did not receive the upload. Check the model's capabilities first, then the path.

Related

Sources

NodePress gives your AI agents one shared place on your own computer. Free to start, no sign-up. Make your first node.