Kimi Guide
Kimi is a broad AI workspace, not just a single chatbot. It currently combines general chat, document work, deep research, website generation, agent workflows, and a coding layer called Kimi Code.
This guide is based on kimi.com/en, Kimi Code docs, and the Kimi API Platform. The product is moving quickly, so feature names and model branding may shift faster than the overall workflow patterns described here.
1. What Kimi is good at​
Kimi currently presents itself around:
- coding,
- multi-step agent work,
- building websites,
- working with documents,
- deep research,
- office-style tasks such as sheets and slides.
If you want one AI tool that covers both knowledge work and developer-adjacent tasks, Kimi is worth a look.
2. The main surfaces​
From the current product navigation and public positioning, the most important Kimi surfaces are:
| Surface | Use it for |
|---|---|
| Chat / main Kimi workspace | General prompting, ideation, planning, and interactive work |
| Websites | Prompt-driven website generation and iteration |
| Docs | Working with uploaded or referenced documents |
| Deep Research | Broader research tasks and synthesis |
| Sheets | Spreadsheet-style workflows |
| Slides | Presentation generation |
| Agent Swarm | Larger tasks broken across agent-style workers |
| Kimi Code | Coding work in terminal, editor, or connected coding tools |
For most people, that means Kimi can be treated as one umbrella product with specialized work modes inside it.
3. How to start using it​
The easiest path is:
- Open Kimi.
- Start with a normal chat task.
- Move into a specialized surface only when the task needs it, for example:
- Websites for prompt-to-site generation
- Docs for document-heavy work
- Deep Research for broader synthesis
- Kimi Code for software tasks
This keeps the tool simple at first and avoids jumping into the more agentic features too early.
4. Where Kimi fits especially well​
Kimi is a good fit when:
- you want one AI workspace instead of many disconnected single-purpose tools,
- your work mixes documents, research, and occasional coding,
- you want to experiment with agent-style workflows without setting up a full developer stack,
- you care about website, slide, or sheet generation alongside chat.
5. Pricing​
What could be verified​
From official public sources on June 26, 2026, I could verify that:
- Kimi has a free plan
- Kimi also has paid membership tiers
- the public pricing UI supports monthly and annual billing cycles
- Kimi Code is included through Kimi membership benefits rather than being sold as a separate personal pay-as-you-go product
What should be checked live​
The exact current end-user membership prices were not exposed cleanly in the public static HTML that could be verified from official sources, and the site appears to render some pricing details dynamically by region/account context.
So if you need the exact current user-facing subscription amount, check the live page directly:
Developer pricing​
If you are looking for API pricing rather than user subscriptions, Moonshot publishes that separately on the Kimi API Platform. The currently published Kimi K2.7 Code prices are:
kimi-k2.7-code:$0.19cache-hit input,$0.95cache-miss input,$4.00output per 1M tokenskimi-k2.7-code-highspeed:$0.38cache-hit input,$1.90cache-miss input,$8.00output per 1M tokens
That matters only if you are building software on top of Kimi rather than simply using the consumer product.
6. Best practices for normal users​
- Start with the standard chat surface before moving into specialized modes.
- Use the specialized surfaces only when they clearly fit the task.
- Treat agentic features as best for bigger tasks, not for every small question.
- If you use Kimi Code, remember it shares quota with the broader membership account.
- Re-check live pricing before making a plan recommendation inside a team.
7. When not to choose Kimi​
Kimi may be a weaker fit if:
- you only need a simple chat assistant,
- you want a tool focused purely on coding and nothing else,
- you need very explicit public pricing tables that stay easy to quote externally,
- you need a compliance decision and cannot rely on fast-moving public product pages alone.