Moonshot AI has launched Kimi Code CLI, an open-source coding agent that runs within the terminal. The device reads and edits code, runs shell instructions, searches information, and fetches net pages. It then chooses its subsequent step based mostly on the suggestions it receives. The challenge is MIT-licensed and lives on GitHub..
Kimi Code CLI is the successor to the older kimi-cli. The brand new agent is written in TypeScript and distributed by way of npm. It really works out of the field with Moonshot AI’s Kimi fashions. It will also be configured to make use of different appropriate suppliers.
What’s Kimi Code CLI
Kimi Code CLI is an AI agent for software program improvement and terminal operations. It will probably implement new options, repair bugs, and full refactors. It will probably additionally discover an unfamiliar codebase and reply structure questions. Batch file processing, builds, and chained take a look at runs are supported too.
The execution mannequin is feedback-driven. The agent plans steps, modifies code, runs exams, and experiences its actions. Learn-only operations run routinely by default. For file edits or shell instructions, the agent asks for affirmation first. This approval circulate retains dangerous actions beneath developer management.
The CLI itself is free and MIT-licensed. Mannequin entry requires Kimi Code OAuth or a Moonshot AI Open Platform API key.
Key Options
Moonshot lists a number of options aimed toward lengthy, centered agent periods:
- Single-binary distribution. One command installs it, with no Node.js setup required.
- Quick startup. Moonshot says the TUI is prepared in milliseconds.
- Goal-built TUI. The interface is tuned for prolonged agent periods.
- Video enter. Drop a display screen recording or demo clip into the chat.
- AI-native MCP configuration. Add and authenticate Mannequin Context Protocol servers by way of
/mcp-config. - Subagents for parallel work. Dispatch built-in
coder,discover, andplansubagents in remoted contexts. - Lifecycle hooks. Run native instructions to gate device calls, audit choices, or set off notifications.
Set up and First Run
Two set up paths exist. The official script wants no pre-installed Node.js.
On macOS or Linux, run the set up script:
curl -fsSL https://code.kimi.com/kimi-code/set up.sh | bash
On Home windows, use PowerShell:
irm https://code.kimi.com/kimi-code/set up.ps1 | iexThe worldwide npm set up requires Node.js 24.15.0 or later:
npm set up -g @moonshot-ai/kimi-codeConfirm the binary, then open a challenge and begin the interactive UI:
kimi --version
cd your-project
kimiOn first launch, sort /login contained in the UI. You may select Kimi Code OAuth or a Moonshot AI Open Platform API key. To run one instruction with out the UI, use kimi -p "your process". To renew the earlier session, add -C.
Use Circumstances
- Understanding a challenge: Ask for an structure overview and a module dependency diagram.
- Implementing a function: Describe the signature, choices, and acceptance standards up entrance.
- Fixing a bug: Give the symptom, replica steps, and anticipated conduct collectively.
- Writing exams and refactoring: Extract repeated patterns, then run exams to verify conduct.
- One-off automation: Analyze logs and output name counts with p50 and p99 latencies.
- Scheduled duties: Ask the agent to set reminders or recurring checks by way of cron.
Plan mode is on the market via Shift-Tab or kimi --plan. It outputs a analysis plan earlier than touching information. For protected batch work, --yolo or /yolo skips approval prompts. The /fork command creates an experimental department you may abandon. The /compact command compresses context to release tokens. For giant investigations, the principle agent can dispatch subagents in parallel.
How Kimi Code CLI Compares
Kimi Code CLI joins a number of established terminal coding brokers. The desk under compares it with three of them. Competitor particulars replicate mid-2026 and might change rapidly.
| Attribute | Kimi Code CLI | Claude Code | OpenAI Codex CLI | Gemini CLI |
|---|---|---|---|---|
| Developer | Moonshot AI | Anthropic | OpenAI | |
| Backing mannequin | Kimi fashions | Claude fashions | GPT-5.3-Codex | Gemini 2.5 Professional |
| Language / runtime | TypeScript | Node.js | Rust | TypeScript |
| Set up | Script or npm (Node.js ≥ 24.15.0) | Native installer or npm | npm / native | npm single binary |
| MCP assist | Sure (/mcp-config) |
Sure | Sure | Sure |
| Subagents | Sure (coder, discover, plan) |
Sure | Sure | No (sequential) |
| Plan mode | Sure (Shift-Tab) |
Sure | Sure | Sure |
| IDE integration | ACP (Zed, JetBrains) | VS Code, JetBrains | VS Code, IDEs | VS Code (Code Help) |
| License | MIT | Proprietary | Open supply | Apache 2.0 |
All 4 brokers assist the Mannequin Context Protocol. They differ on backing mannequin, language, license, and orchestration. Kimi Code CLI and Codex CLI each ship native subagents. Gemini CLI runs duties sequentially with out subagent assist.
Key Takeaways
- Kimi Code CLI is an MIT-licensed terminal coding agent from Moonshot AI.
- It’s written in TypeScript and installs by way of script or npm.
- Constructed-in
coder,discover, andplansubagents run in remoted contexts. - MCP servers are configured conversationally via
/mcp-config, not uncooked JSON. - It succeeds kimi-cli and migrates present configuration and periods.
Marktechpost’s Visible Explainer
Kimi Code CLI · Information
01 / 09
Try the GitHub Repo right here. Additionally, be happy to comply with us on Twitter and don’t overlook to affix our 150k+ ML SubReddit and Subscribe to our Publication. Wait! are you on telegram? now you may be a part of us on telegram as properly.
Have to associate with us for selling your GitHub Repo OR Hugging Face Web page OR Product Launch OR Webinar and so on.? Join with us
Michal Sutter is an information science skilled with a Grasp of Science in Information Science from the College of Padova. With a stable basis in statistical evaluation, machine studying, and knowledge engineering, Michal excels at remodeling complicated datasets into actionable insights.
