China AI Hub
Agents / Kimi Code
Kimi Code
Moonshot AI's terminal AI coding agent ('The Starting Point for Next-Gen Agents'), successor to the deprecated kimi-cli. CLI (TypeScript, MIT), VS Code extension and Desktop app; reads/edits code, runs shell commands, searches files, fetches web pages, plans and adjusts actions autonomously. Includes subagents, MCP, Kimi Computer Use, browser control (WebBridge/Browser Extension) and multimodal input (text, images, video). Billed under Kimi membership.
Kimi Code is a coding agent by moonshot-ai, built on kimi-k3, kimi-k2.7-code, kimi-k2.7-code-highspeed, deployable as both a cloud service and self-hosted, and released open source.
| Key fact | Value |
|---|---|
| Company | moonshot-ai |
| Type | coding |
| Framework | TypeScript terminal agent (pi-tui); succeeds the deprecated Python kimi-cli |
| Deployment | both |
| Open source | Yes (MIT) |
| GitHub | https://github.com/MoonshotAI/kimi-code |
| Documentation | https://moonshotai.github.io/kimi-code/en/ |
| Pricing | Included with Kimi membership, Plus and above (Adagio free tier has no coding quota; Plus $15/mo, Pro $31/mo, Max $79/mo, Ultra $159/mo). All clients share one quota with rolling 5-hour window and monthly total. Open Platform API pay-as-you-go: kimi-k3 $3.00/M input / $15.00/M output; kimi-k2.7-code $0.95/$4.00; kimi-k2.7-code-highspeed $1.90/$8.00. |
Capabilities
| Capability | Supported |
|---|---|
| Tool calling | Yes |
| Browser use | Yes |
| Computer use | Yes |
| MCP support | Yes |
| Memory | Unknown |
| Planning | Yes |
| Multi-agent | Yes |
| API | Yes |
Use cases
- Understanding unfamiliar codebases and implementing features
- Bug fixing, writing tests, refactoring
- Batch file processing
- Long-horizon whole-repo work via 1M-token context (K3)
- Video-input tasks (screen recording to code)
- Scheduled and background tasks
- IDE-driven sessions via VS Code extension or ACP (Zed, JetBrains)
Limitations
- Runs on the local machine with approval gates; no dedicated OS-level sandbox engine
- kimi-cli predecessor deprecated and being wound down (auto-migrated on install)
- Computer Use (macOS) needs Accessibility + Screen Recording permissions; Windows version may briefly take over mouse/keyboard
- Third-party tools must keep the real client identifier (User-Agent tampering restricted)
- Cloud inference only (managed endpoints api.kimi.com/coding/v1 and api.kimi.ai/coding/v1); not self-hostable as a model service
- Windows install requires Git for Windows
- API keys shown only once (max 5); quota shared across devices; devices inactive >30 days unbound
Kimi Code is Moonshot AI’s terminal AI coding agent and the successor to the deprecated kimi-cli. It ships as a TypeScript CLI (MIT), a VS Code extension and a Desktop app, and can read and edit code, run shell commands, search files, fetch web pages, and plan and adjust actions autonomously. Subagents, MCP, Kimi Computer Use, browser control and multimodal input (text, images, video) are built in.
It runs on Kimi K3, Kimi K2.7-Code and Kimi K2.7-Code-HighSpeed — K3’s 1M-token context enables long-horizon whole-repo work.
Billing follows Kimi membership (Plus and above); the Open Platform API is pay-as-you-go at $3.00/M input / $15.00/M output for kimi-k3. See the Moonshot AI profile.
Underlying models
Sources
Data interpretation
| Field | Value | Evidence type |
|---|---|---|
| Company | moonshot-ai | Official |
| Type | coding | Official |
| Underlying models | kimi-k3, kimi-k2.7-code, kimi-k2.7-code-highspeed | Official |
| Open source | Yes (MIT) | Official |
| Deployment | both | Official |
| Capabilities | Tool calling, Browser use, Computer use, MCP support, Planning, Multi-agent, API | Official |
Evidence types: Official = vendor documentation or official pages. See the sourcing policy.
What is Kimi Code?
Moonshot AI's terminal AI coding agent ('The Starting Point for Next-Gen Agents'), successor to the deprecated kimi-cli. CLI (TypeScript, MIT), VS Code extension and Desktop app; reads/edits code, runs shell commands, searches files, fetches web pages, plans and adjusts actions autonomously. Includes subagents, MCP, Kimi Computer Use, browser control (WebBridge/Browser Extension) and multimodal input (text, images, video). Billed under Kimi membership.
Which models does Kimi Code use?
Kimi Code uses kimi-k3, kimi-k2.7-code, kimi-k2.7-code-highspeed.
Is Kimi Code open source?
Yes — licensed under MIT.
How much does Kimi Code cost?
Included with Kimi membership, Plus and above (Adagio free tier has no coding quota; Plus $15/mo, Pro $31/mo, Max $79/mo, Ultra $159/mo). All clients share one quota with rolling 5-hour window and monthly total. Open Platform API pay-as-you-go: kimi-k3 $3.00/M input / $15.00/M output; kimi-k2.7-code $0.95/$4.00; kimi-k2.7-code-highspeed $1.90/$8.00.
Where does China AI Hub get its Kimi Code data?
From 4 sources, last verified 2026-09-20.