This week, AI assistants connected to OKiDO over MCP gained a clearer view of your operational context. They can now identify the files attached to your work and open the images they need to understand screenshots, diagrams, and other visual information.
That closes an important gap between what you see in OKiDO and what an assistant can use. Instead of encountering an inaccessible link or having to interpret the editor’s underlying document structure, assistants can work with attachments through dedicated, permission-aware tools.
AI assistants can now see images inside your work
An MCP-connected assistant can now retrieve the actual image behind an upload in OKiDO. The new get_upload tool lets the assistant open pictures embedded in task descriptions, task comments, and wiki documents.
Previously, an assistant could encounter a reference to an uploaded image but receive only a link it could not open. That meant important context contained in a screenshot or diagram remained visible to you but unavailable to the assistant. It could read the surrounding text, but not the visual evidence that text referred to.
With get_upload, you can ask an assistant to work with visual material that is already part of the task or document it is reviewing. For example, it can inspect a screenshot included with a reported issue, refer to a process diagram in a wiki document, or use an image posted in a task comment as additional context.
This improves several common ways of working:
Reviewing reported issues: An assistant can inspect a screenshot attached to a task instead of relying only on the written description.
Understanding documented processes: It can retrieve diagrams and other inline images embedded in wiki documents.
Following conversations: It can open an image added to a task comment when that image provides context for the latest update.
Working from complete evidence: It can combine written instructions with the visual material included alongside them.
You no longer need to duplicate the contents of an image in the surrounding text just to make it available to an assistant. The operational context can stay where your team naturally records it: inside the task, comment, or document.
Attachments are easier for assistants to discover
Before an assistant opens an image, it first needs to know that the image exists. The get_task and get_document MCP tools now return an attachments array containing every embedded file and inline image associated with the task or document.
Each attachment entry includes its name, type, and size. This gives the assistant a straightforward inventory of the available files without requiring it to parse the editor’s document format or guess which references represent uploads.
The same improvement now applies to task discussions. list_task_comments includes an attachments array for each individual comment, making it clear which files belong to each part of the conversation. An assistant can distinguish between an image in the original task description and one added later in a specific comment.
Assistants can follow a more reliable workflow
With structured attachment data, an assistant can now follow a consistent sequence:
Read the task, document, or task comments.
See which files and inline images are attached.
Check the name, type, and size of each attachment.
Retrieve the relevant image with
get_uploadwhen visual context is needed.
This matters when a task contains several uploads or when a long comment thread accumulates supporting material over time. Rather than treating attachments as opaque links, the assistant can identify what is available and select the item relevant to your request.
It also makes assistant behavior easier for you to direct. You can ask it to review the screenshot in a task, compare images from two comments, or inspect the diagram attached to a document, knowing it has a structured way to locate those files.
Image access follows the work it belongs to
When an assistant uses get_upload, OKiDO checks the file against the task or document it belongs to. The assistant can open an image only when its user can already view the parent work.
This applies across the places where uploads can appear, including tasks and documents inside private projects. Being able to identify or request an upload does not give the assistant broader access to the surrounding work. The visibility of the task or document remains the deciding factor.
That relationship is important because attachments rarely stand alone. A screenshot may contain details that make sense only within a particular task, while a diagram may document a process intended for a specific project. By tying file access to the parent item, OKiDO keeps the assistant’s visual context aligned with the operational context already available to its user.
The result is consistent behavior whether the assistant is reading text or retrieving an image. If the user can work with the task or document, the assistant can use its relevant attachments. If the parent work is not available to that user, its embedded images are not available through get_upload either.
Put visual context to work
This week’s changes make visual context a usable part of MCP-based workflows. Assistants can discover attachments in tasks, documents, and comments, understand basic details about each file, and retrieve relevant images instead of stopping at an unusable link.
Try it by connecting your MCP assistant to a task or wiki document that contains a screenshot or diagram. Ask it to list the available attachments, identify the relevant image, and use that visual context when reviewing the work.
