★ Featured
Kimi
A brief turns into the finished file: Kimi runs the research, builds the deck or the spreadsheet, and keeps working on the desktop and in the terminal.
Chatbots & Assistants
LLM
Open source
AI-native
AI Agents
Disclosure: this is an affiliate link. We may earn a commission.
Pricing
Freemium
Free plan available, paid plans to go further.
Categories
Chatbots & Assistants · AI Agents
Platforms
Web, iOS, Android, macOS, Windows, API
Languages
Multilingual
Last updated
July 27, 2026
Overview
A quick question, a large dossier and a project that runs for hours all start in the same input box and come back as a finished file. Kimi answers in the chat, then builds the object itself: a deck, a spreadsheet, a sourced research report or a website ready to publish. Agent Swarm splits heavy work across sub-agents running in parallel, which shortens a literature review or a sweep of hundreds of sources. The Kimi Work desktop app goes one level down: it reads local folders, drives the browser on the user's behalf and fires tasks on a schedule, with the chosen permission level deciding whether it stops before each sensitive action. Kimi Code carries the same engine into the terminal and the editor. Among AI chatbots and assistants, Kimi stands out for openness: Moonshot AI publishes the weights of its models, which other hosts serve and teams can run in house.
Key features
- Kimi K3 takes up to one million tokens in a single request and understands images natively.
- Slides, spreadsheets, documents and deployable websites generated from a brief written in plain language.
- Deep Research reads through a large set of pages and returns a sourced report with interactive charts.
- Agent Swarm hands pieces of a task to sub-agents that work at the same time, then merges what they bring back.
- The Kimi Work desktop app opens local folders, drives the browser through WebBridge and runs scheduled tasks.
- Kimi Code works from the terminal or the editor: it reads the repository, edits files and runs commands.
Best use cases
- A financial analyst queries the market data built into the desktop app and leaves with an annotated spreadsheet.
- A doctoral student drops in forty PDFs and gets back a paginated literature review, references included.
- A developer hands Kimi Code a rewrite spread across several files, straight from the terminal.
- A consultant turns a scoping note into an illustrated deck to present to a client.
- A freelancer schedules a daily brief that compiles the morning news for their sector.
- A technical team plugs the model into its own products through the API and pays per token.
Pros
- Moonshot AI publishes the weights of its Kimi models, so other providers serve them and teams can host them in house.
- The free plan has no time limit and still runs the most advanced model, one agent task at a time.
- The desktop app runs either in ask permission mode, which pauses on every sensitive action, or in full access mode, which lets it work uninterrupted.
- The API caches context automatically, bringing reused input tokens down to a tenth of the standard rate.
