What I want Cursor cloud agent to be able to do is view images posted in Slack threads that someone then uses @Cursor in. This is because our team often post screenshots of bugs in the threads, which are required for Cursor to fix the issue.
I dug into this and when I trigger Cursor from Slack, it doesn’t pass the thread ID (or even the channel) in the context. In the agent I see.
<slack_context>
<slack_channel>
Channel: #edu-rbwe-eng
Channel purpose: xxxx
</slack_channel>
<slack_thread>
The user shared the following Slack thread for additional context
[3/18/2026, 5:05:35 AM] yyy: Hi xxxthese are the med priority enhancement / bugs that we can work on for the next sprint:
Canvas background - can we come up with a responsive design that allows for a larger image to be uploaded so that the image always fills up the entire screen?
Chip UI is not the same as Figma design
Can we reduce this pop-up duration to 1.5 secs
Image of Al & Min is chopped off
There is a gap between the two images, diff from Figma design
Can we remove this button [Image 1] [Image 2] [Image 3] [Image 4] [Image 5] [Image 6]
[3/18/2026, 5:49:45 AM] Robin Weston: I reckon Cursor can one shot most of these
[3/18/2026, 5:51:04 AM] Robin Weston: @Cursor can you see the app screenshot images posted in this thread? Don’t write any code. use edu-rbwe repo
</slack_thread>
</slack_context>
My fallback plan was to plug in a Slack MCP or CLI into Cursor so it could pull back the images itself. However, no thread ID is passed. All it has to go on is the channel name. So it could probably find the channel by name, then search for threads with matching content to recent messages, but that is quite messy and prone to failure.
Steps to Reproduce
Create a Slack thread. Add some images to it. Then @Cursor
Expected Behavior
I would expect there to be a way for Cursor to pull back images from a Slack thread that it was commented in using @Cursor
Operating System
MacOS
Version Information
Cloud Agents
For AI issues: add Request ID with privacy disabled
Hey, thanks for the detailed report and for digging in so deeply.
A couple quick clarifications:
Images really are passed to the agent. The placeholders [Image 1], [Image 2], etc. that you see in the thread context aren’t just text. Each one maps to a real image that gets downloaded from Slack and sent to the model. So the agent should be able to see them.
If the agent replied like it couldn’t see the images, that’s more likely a model behavior issue, not a data pipeline issue. Can you share what the agent actually said? Did it explicitly say it can’t see images, or did it just not reference them at all?
Thread or channel IDs in the agent context. You’re right, those aren’t passed through right now. The system uses them internally to route replies back to Slack, but the agent and MCP tools can’t access them.
As a temporary workaround, since images are already being passed, you can try making the prompt more explicit, like: “look at the images posted in this thread and describe what you see” to nudge the model to use the visual data.
Let me know what the agent’s actual response was about the images. That’ll help figure out whether it’s a model issue or something else.
The user is asking me to look at an image. However, I don’t see any image attached to this message. Let me check the context again…
Looking at the slack_context, there’s a mention of “[Image 1]” in the thread, but I don’t actually have access to that image in the current context. The user is asking “what do you see in this image?” but I don’t have an image to analyze.
I should let the user know that I don’t see an image attached, and ask them to share it, or I should check if there’s a way to access it. However, based on the Slack thread context, it seems like the issue is about “zoomed in font looks really weird on GPS wayfinding” and there’s discussion about applying a CSS fix globally across the app.
Hey Robin, this is really helpful. Thanks for running that test and sharing the agent link.
Based on the agent’s reasoning, I don’t see any image attached to this message, it’s clear the images aren’t reaching the model even though they’re being processed on our side. This looks like a pipeline bug, not a model behavior issue, so I take back what I said earlier.
I’ve flagged this with the team along with your agent link so they can trace exactly where the images are getting dropped. I don’t have a timeline for a fix, but your detailed repro makes this much easier to investigate.
For now, the workaround is to paste or link the images directly in the cloud agent follow-up instead of relying on the Slack thread context. Not ideal, I know.
I can see that the Slack thread images are now included in the thread context passed to the agent. Thank you.
However, the agent doesn’t actually have direct access to the images and so can’t manipulate them as files. This would be helpful to us as we want to use Cursor to upload screenshots to the bug ticket it has created for us.
About accessing images as files, that’s not supported yet. Right now, images from Slack are sent to the model for vision, but the agent can’t download them or attach them anywhere as files.
I’ve forwarded this request to the team. There’s no timeline yet, but the use case is clear, bug screenshots from Slack straight into tickets.
As a workaround for now, you can ask the agent to describe the bug based on the screenshot, and then attach the images to the ticket manually.
