|

Moonshot AI Launches Kimi Work, a Local Desktop Agent Reportedly Running on Kimi K2.6 With a 300-Sub-Agent Agent Swarm

Moonshot AI has launched Kimi Work, an AI agent that runs on your personal desktop. The Beijing-based AI entity introduced it this week together with downloads for macOS and Windows. Kimi Work reads native information, drives your actual browser, and runs scheduled duties. It targets information staff whose bottleneck is entry to information and reside classes.

Most agent instruments of the previous two years ran within the cloud. You sort a aim, a distant server spins up a sandbox, and a hosted browser acts. Kimi Work runs domestically as a substitute, reaching information and classes you already use.

What is Kimi Work?

Kimi Work is a downloadable software, not a net chat. You give it objectives in plain language, and it acts on your machine. Independent neighborhood mentions report that it runs on Kimi K2.6, Moonshot’s flagship mannequin.

K2.6 is an open-weight Mixture-of-Experts mannequin launched on April 20, 2026. It prompts about 32 billion parameters per token. It carries a 256K-token context window for lengthy, multi-step work.

How Kimi Work Operates

Four constructing blocks outline the product. Knowing them helps you motive about what it may possibly do.

  1. Agent Swarm: Kimi Work can run many sub-agents in parallel on your machine. According to Moonshot launch, the swarm scales to 300 sub-agents. The system splits a activity into elements, then coordinates the outcomes. K2.6’s swarm is documented as much as 4,000 coordinated steps.
  2. WebBridge: This browser extension lets the agent use a browser like a particular person. It searches, scrolls, extracts information, and fills varieties throughout tabs. Because it makes use of your actual session, it inherits your current logins and cookies.
  3. Cron scheduling engine: A built-in scheduler runs jobs on a day by day, hourly, or conditional foundation. Per Moonshot, triggers embrace LLM agent calls and Python or shell scripts. A “Keep Computer Awake” toggle retains in a single day jobs from stalling.
  4. Local information and code: The agent reads folders you mount and runs Python within the background. According to Moonshot launch, authentic information keep in place until you approve a change.

The desktop app additionally ships finance-specific information. It is pre-integrated with market information for A-shares, Hong Kong shares, and US equities. According to Moonshot launch, this removes the necessity for customized API setup. Finished analysis can convert into EnergyPoint decks or Excel sheets.

Use Cases With Examples

  • Document triage: Point the agent at a folder of quarterly PDFs. Ask it to summarize them into one doc, protecting originals intact. The swarm assigns one reader per file, then merges findings.
  • Web information assortment: Tell WebBridge to drag historic costs for 3 tickers. It opens your browser, units the date vary, and extracts the tables. Python then normalizes columns and writes an Excel workbook.
  • Scheduled briefings: Define a 7:00 AM job within the cron engine. Each morning it gathers headlines and drafts a markdown briefing. With “Keep Computer Awake” on, the job survives in a single day.
  • Office era: Ask for a quick market-brief deck after a analysis move. The agent drafts sections in parallel and renders native slides.

Kimi Work vs Cloud Agents

The core distinction is the place the agent runs and what it may possibly attain. The desk compares Kimi Work in opposition to a typical cloud agent.

Dimension Kimi Work (native) Typical cloud agent
Execution location Your desktop Vendor servers
File entry Mounts your native folders Uploaded or sandboxed information
Browser Your actual, logged-in browser through WebBridge Hosted digital browser
Scheduling Built-in cron engine Often exterior or restricted
Underlying mannequin Kimi K2.6, reported Vendor’s hosted mannequin
Setup Install app, grant folder entry Zero-install, open a tab
Security duty Falls on the consumer Falls on the seller

Neither method wins outright. Local execution retains information on your machine and reaches actual information. Cloud execution trades that management for zero-setup comfort and managed security.

Scheduling: The Cron Engine in Practice

Kimi Work is pushed by pure language, not a public API. Its scheduler is a cron engine, so it accepts customary cron schedules. The 5 fields are: minute, hour, day-of-month, month, and day-of-week.

# Standard cron schedules the engine understands
0 7 * * *      # day by day at 07:00
0 * * * *      # each hour, on the hour
30 8 * * 1-5   # 08:30 on weekdays solely (Mon-Fri)
0 0 1 * *      # 00:00 on the primary day of every month

You pair a schedule with a plain-language activity. A day by day briefing job reads like this.

Schedule: 0 7 * * *   (day by day at 07:00)
Task:     "Draft right this moment's market briefing and reserve it to
           ~/KimiWorkarea/briefing.md. Ask earlier than writing."

The approval gate then applies to that write, and to any net motion.

Key Takeaways

  • An “Ask earlier than appearing” gate, with YOLO mode off, prompts earlier than any file write.
  • Kimi Work is a native desktop agent for macOS (Apple silicon) and Windows.
  • An Agent Swarm runs as much as 300 sub-agents in parallel on your machine.
  • WebBridge drives your logged-in browser; a built-in cron engine runs scheduled jobs.
  • It reads native folders and runs Python, protecting originals until you approve modifications.

Marktechpost’s Interactive Explainer









Check out the Details hereAlso, be happy to comply with us on Twitter and don’t overlook to hitch our 150k+ ML SubReddit and Subscribe to our Newsletter. Wait! are you on telegram? now you can join us on telegram as well.

Need to associate with us for selling your GitHub Repo OR Hugging Face Page OR Product Release OR Webinar and so on.? Connect with us

The put up Moonshot AI Launches Kimi Work, a Local Desktop Agent Reportedly Running on Kimi K2.6 With a 300-Sub-Agent Agent Swarm appeared first on MarkTechPost.

Similar Posts