Kimi Work - The AI desktop for knowledge work
Kimi Work is a desktop agent for knowledge work. It connects to local files, uses WebBridge for browser automation, runs scheduled tasks, coordinates agent swarms, creates PPT/Excel/Word/PDF outputs, and includes native finance data tools.
Replies
Flowtica Scribe
Hi everyone!
Kimi Work brings Kimi to the desktop as a local agent. It has access to your files, can run browser automation via WebBridge, and supports running many agents in parallel for bigger tasks.
It also includes a scheduler, so you can set up recurring jobs and let Kimi run them in the background.
For heavier jobs, it can spin up the agent swarm — up to 300 agents working in parallel — then turn the result into Excel, PPT, Word, or PDF.
@zaczuo Interesting combination of features, especially the ability to analyze large batches of files alongside real-time web search. The enhanced image understanding also caught my eye. Curious, which use case are users gravitating toward most right now: research, coding, or content creation?
a desktop agent that can actually create finished PPT and Excel files instead of just talking about them is what I've been waiting for. most AI tools stop at "here's a draft in chat" and then you spend 20 minutes copy-pasting into the actual document. the scheduled recurring tasks in the background is a nice touch too
viktor.com
cool idea!
Mailwarm
How do you handle permissions around local files and web actions so it doesn’t accidentally do something risky?
This feels useful for people who jump btw files, browser tabs and docs all day. I like that it is not just chat but closer to a work desktop. How well the scheduled tasks work with local files
DIY UX Test
A dedicated desktop for knowledge work rather than another chat tab is the right instinct — context-switching is what kills most AI workflows. Is the K2 model running locally for any of it, or all server-side?
Knowledge work tools often start as search systems but eventually become context systems.
Have you found users spend more time retrieving information they've already seen, or building new knowledge structures that persist over time?
Nice! Like a part of the cmdop.com
scheduled tasks feature is where desktop agents either become genuinely useful or quietly cause problems. a task that runs while you're not watching needs clear boundaries around what it's allowed to do, especially when it has file access and browser automation. what does the review and approval flow look like for scheduled agent actions and is there a way to see exactly what happened after a scheduled task ran