Can Codex CLI see the images?

[neznayer]

I have a MPC server with Playwright that can take screenshots. Is there a way for the model to "see" the screenshots?
Because here https://help.openai.com/en/articles/11096431-openai-codex-cli-getting-started
it is mentioned that it is a multimodal. It is obsolete?

[aehlke]

They've said it's coming "soon"

[pidgeon777]

It is now supported. Just copy and paste the path of the image to the codex input form.

[aehlke]

Can MCPs provide screenshots to it via path?

[pidgeon777]

I don't know this, but would be interesting.

[Nemo64]

I don't think so.

I use an MCP for my ticket system and there are often screenshots in them.

It does say <image content> but it does not appear as if it gains any knowledge from those images.

It also very often just completely eats up the context which makes me think that it just pasts the base64 content in the context.

Claude Code has no issue looking at those images and even reliably reading the text in them, which I'm always impressed by.

[aehlke]

I filed #3741

[eugkabanov]

it's now supported. drag&drop or control ⌃ + V for clipboard images works.

[gommo]

Really?

[harsh62]

Pasting from clipboard is not working for me atleast.

[YoloMark98]

working but sometimes.. many times it fails

[deepakness]

Pasting images is working for me by using control + v. Attaching a screenshot for the same as well:

[dante01yoon]

Is it working? I have no confidence because it only gives preview how size it is