Kimi-CLI-X
A terminal coding agent forked from Kimi CLI, with lean prompts, Python scripting and support for dozens of API providers.
- 111 ★ on the repository
- Last commit
Broad access
Why this level
- The agent runs shell commands, executes scripts and edits project files.
- Provider API keys are stored in the user config.
Install
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/kimi-cli-xDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The agent runs shell commands, executes scripts and edits project files.
- Provider API keys are stored in the user config.
This is third-party code. Review the repository files before installing.
What it does
Kimi-CLI-X, or KimiX, is a deep rework of Kimi CLI, a terminal AI agent for coding. The authors compressed the initial prompts and tool descriptions to roughly two thousand tokens, hardened permission checks and output handling, and simplified subprocesses, sub-agents and background tasks. The main difference is Python scripting: you write an ordinary script with loops and conditions that calls `prompt()` and tools, and get a repeatable process instead of a chat. New capabilities include interactive Bash and Powershell sessions with pattern waiting, Docx and PDF conversion, Python script execution, headless-browser `fetch_url` and best-of-N parallel sampling in isolated working copies. The agent connects to Kimi models and to OpenAI, Anthropic, Google and other providers through a user config.
Who it is for. Developers who want a standalone terminal agent with programmable workflows and a free choice of model.
Good fit when
- You need to process files in bulk: a loop over documentation with a request to the agent for each.
- You want the best of several parallel runs of one task, with verification.
- The agent must hold interactive command-line sessions and wait for output.
- You need to plug in your own provider with an OpenAI- or Anthropic-compatible interface.
Not a fit when
- You need an agent plugin for an editor you already use rather than a separate terminal tool.
- Running an agent with shell access on the work machine is forbidden.
- You need an agent with guaranteed support and a license: the repository declares none.
Example request
Go through every file in docs, find mismatches with the latest commits, update the descriptions and show the diff.Limitations
Installation runs `python install.py`, which syncs dependencies, downloads the native runtime and registers tools, so network access and write permission are needed. You need an API key for the chosen model, config templates live in `docs/` and a config is passed with the `--config` flag. The agent runs shell commands and edits files, so set the permission mode deliberately. The repository declares no license, and the main docs are bilingual.
How to disable. Delete the clone directory and the configuration created by `install.py`. API keys live in your config and are revoked with the provider separately.
FAQ
How do I install KimiX?
Run `python install.py` from the repository root. The installer syncs dependencies, downloads the native runtime and registers tools.
Does the agent work only with Kimi?
No. You can plug in OpenAI, Anthropic, Google and other APIs: config templates are in `docs/` and you start with `kimix --config=xx.json`.
What are Python workflows?
A script with `from kimix import *` that calls `prompt()` per file in a loop, giving a reproducible batch process.
How does the agent handle long tasks?
Context is indexed with BM25, old messages are compacted automatically and a turn cannot end while todos are unfinished or edits lack a check.
What is parallel sampling?
The `parallel_sample` mode runs the same task N times in isolated copies, picks the best by self-evaluation or majority vote and applies it after verification.
Discussion
No questions or reviews yet.
Related
CLI toolsKimi Code is Moonshot AI's terminal coding agent: it reads the project, edits files and runs commands. Installs with one script, no Node.js needed.
Meta_Kim guides an agent from idea to verified result: it gathers context, plans, assigns tasks and writes lessons back.
A terminal coding agent built on Codex, tuned for low-cost open models like Kimi K3, DeepSeek and Qwen
A teaching terminal coding agent in Python, built as a "nanoGPT for agents": read it, take it apart and fork it
An MCP server that hands large-repo analysis and independent diff review to Kimi Code so Claude spends fewer tokens reading files.
A CLI agent that works with project files and shell commands, lets you choose a model provider and connects external MCP tools.
Comparison with similar tools
Cards that share the main search query, and how they differ.
README in short
The project brings programmability and long-run resilience to Kimi CLI: context compaction, loop protection and todo reminders. Tutorials live in `docs/tutorials`.