← Back to Skills Marketplace
jack-yang-ai

KimiClaw

by Jack-Yang-ai · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ⚠ suspicious
141
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install kimiclaw
Description
KimiClaw: Power your OpenClaw with Kimi K2.5 — the free, Anthropic-compatible coding model. One config change to run Claude Code, spawn coding agents, or cha...
README (SKILL.md)

🦞 KimiClaw — Power Your OpenClaw with Kimi K2.5

KimiClaw:用 Kimi K2.5 驱动你的 OpenClaw

Kimi Code (K2.5) is fully compatible with Anthropic Messages API. One config change and your OpenClaw runs on Kimi — Claude Code, coding agents, chat, everything.

Kimi Code (K2.5) 完全兼容 Anthropic Messages API。改一行配置,你的 OpenClaw 就跑在 Kimi 上 — Claude Code、编码 Agent、聊天,全都行。


Get Your API Key / 获取 API Key

👉 Kimi Code Console

Create a key → format: sk-kimi-...

创建密钥 → 格式:sk-kimi-...


1. OpenClaw Provider Setup / OpenClaw 配置

Add to openclaw.json:

openclaw.json 中添加:

// providers:
{
  "id": "kimi",
  "type": "anthropic",          // Kimi speaks Anthropic protocol / Kimi 用 Anthropic 协议
  "baseUrl": "https://api.kimi.com/coding",
  "apiKey": "sk-kimi-..."
}

// models (or agents.defaults.models):
{
  "kimi/kimi-k2.5": {
    "alias": "Kimi K2.5"
  }
}

Now use it anywhere / 随处可用:

/model kimi/kimi-k2.5          # Switch in chat / 聊天中切换

Or set as agent default / 或设为 Agent 默认模型:

"model": "kimi/kimi-k2.5"

2. Claude Code CLI

export ANTHROPIC_BASE_URL="https://api.kimi.com/coding"
export ANTHROPIC_API_KEY="sk-kimi-..."

claude                          # Interactive / 交互模式
claude --print "Your prompt"    # One-shot / 单次执行

Claude Code auto-appends /v1/messages. No other changes.

Claude Code 自动拼接 /v1/messages,无需其他改动。


3. Spawn Coding Agent / 启动编码 Agent

OpenClaw spawns Claude Code with Kimi backend:

OpenClaw 用 Kimi 后端启动 Claude Code:

sessions_spawn(
    runtime="acp",
    task="Refactor auth module to use JWT",
    env={
        "ANTHROPIC_BASE_URL": "https://api.kimi.com/coding",
        "ANTHROPIC_API_KEY": "sk-kimi-..."
    }
)

Or configure globally in openclaw.json — every ACP spawn uses Kimi by default.

也可在 openclaw.json 全局配置,每次 spawn 默认走 Kimi。


API Reference / API 参考

Property / 属性 Value / 值
Base URL https://api.kimi.com/coding
Endpoint / 接口 https://api.kimi.com/coding/v1/messages
Auth x-api-key: sk-kimi-...
Version anthropic-version: 2023-06-01
Model (request / 请求) kimi-k2.5
Model (response / 响应) kimi-for-coding
Streaming / 流式 "stream": true → SSE

Quick Test / 快速验证

curl -s https://api.kimi.com/coding/v1/messages \
  -H "x-api-key: sk-kimi-..." \
  -H "anthropic-version: 2023-06-01" \
  -H "content-type: application/json" \
  -d '{"model":"kimi-k2.5","max_tokens":1024,"messages":[{"role":"user","content":"Hello"}]}'

Python (no dependencies / 纯标准库)

import json, urllib.request

req = urllib.request.Request(
    "https://api.kimi.com/coding/v1/messages",
    data=json.dumps({
        "model": "kimi-k2.5",
        "max_tokens": 4096,
        "messages": [{"role": "user", "content": "Hello"}]
    }).encode(),
    headers={
        "Content-Type": "application/json",
        "x-api-key": "sk-kimi-...",
        "anthropic-version": "2023-06-01",
    },
)
with urllib.request.urlopen(req, timeout=120) as resp:
    print(json.loads(resp.read())["content"][0]["text"])

Gotchas / 注意事项

⚠️ EN 中文
Model name Request: kimi-k2.5 → Response: kimi-for-coding. Don't assert on response. 请求发 kimi-k2.5,响应返回 kimi-for-coding,不要断言响应模型名。
Format Anthropic only (/v1/messages). OpenAI format (/v1/chat/completions) → 404. 仅支持 Anthropic 格式,OpenAI 格式返回 404。
moonshot.cn api.moonshot.cn is a different product — different models, different auth. api.moonshot.cn 是另一个产品,模型和认证都不同。
Timeout Set ≥120s for complex prompts. 复杂提示词设 ≥120 秒。
Provider type Must be "type": "anthropic" in OpenClaw config. OpenClaw 配置里必须写 "type": "anthropic"
Usage Guidance
This skill appears to be what it says (instructions to use Kimi K2.5 via an Anthropic-compatible API), but exercise caution before providing API keys or modifying configs: 1) The registry metadata did not declare any required env vars though the instructions clearly tell you to set ANTHROPIC_API_KEY / ANTHROPIC_BASE_URL or to put an apiKey in openclaw.json—confirm that behavior with the skill author. 2) Verify the provider domain (https://api.kimi.com and https://www.kimi.com) and ownership before using a real key—prefer to test with a limited-scoped or expendable API key first. 3) Don’t put high-privilege secrets or other service credentials into openclaw.json or wide-scope env vars; create a key with limited quota/permissions if possible. 4) Confirm expected request/response formats in a safe test (curl) so you don’t inadvertently send sensitive data. 5) Because there is no homepage or maintainer info in the registry metadata, consider requesting provenance (source repo or publisher contact) before relying on this skill in production. If the publisher provides a repo or official docs, re-check that they match the endpoints and headers used in SKILL.md.
Capability Analysis
Type: OpenClaw Skill Name: kimiclaw Version: 1.0.0 The skill bundle is a configuration guide for integrating the Kimi K2.5 LLM (via api.kimi.com) into OpenClaw. It contains standard documentation, JSON configuration snippets, and benign Python/shell examples for API interaction in SKILL.md. No malicious intent, data exfiltration, or harmful instructions were detected.
Capability Assessment
Purpose & Capability
The SKILL.md content is consistent with the name/description: it instructs how to point OpenClaw and Anthropic-compatible tooling at Kimi's coding API. However, the registry metadata declares no required credentials/env vars while the instructions clearly show using an API key (sk-kimi-...) and environment variables (ANTHROPIC_API_KEY / ANTHROPIC_BASE_URL). That mismatch is an implementation/documentation inconsistency.
Instruction Scope
The runtime instructions are narrowly scoped to configuring OpenClaw, setting environment variables, and calling the Kimi API endpoints. There are no instructions to read unrelated local files, harvest system config, or transmit data to unexpected endpoints beyond the documented api.kimi.com domain. Examples do instruct using an API key and show curl/python usage for requests (normal for this kind of skill).
Install Mechanism
This is an instruction-only skill with no install spec and no code files, which minimizes install-time risk. Nothing is downloaded or written by the skill instructions themselves.
Credentials
The skill requires an API key in practice (examples show sk-kimi-... and recommend ANTHROPIC_API_KEY), but the registry metadata declares no required environment variables or primary credential. Requesting an API key for the service the skill integrates with is reasonable, but the absence of that declaration is a discrepancy and reduces transparency. Also, the skill will cause the user to place an API key into openclaw.json or env vars — users should ensure that key is scoped/limited and that they trust the target domain.
Persistence & Privilege
always is false and the skill does not request persistent privileges or attempt to change other skills' configurations. It only gives instructions for user configuration of OpenClaw (expected for a provider integration).
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install kimiclaw
  3. After installation, invoke the skill by name or use /kimiclaw
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
KimiClaw v1.0: Power your OpenClaw with Kimi K2.5. One-click setup for OpenClaw provider, Claude Code CLI, and ACP coding agents. Bilingual EN+CN.
Metadata
Slug kimiclaw
Version 1.0.0
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 1
Frequently Asked Questions

What is KimiClaw?

KimiClaw: Power your OpenClaw with Kimi K2.5 — the free, Anthropic-compatible coding model. One config change to run Claude Code, spawn coding agents, or cha... It is an AI Agent Skill for Claude Code / OpenClaw, with 141 downloads so far.

How do I install KimiClaw?

Run "/install kimiclaw" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is KimiClaw free?

Yes, KimiClaw is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does KimiClaw support?

KimiClaw is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created KimiClaw?

It is built and maintained by Jack-Yang-ai (@jack-yang-ai); the current version is v1.0.0.

💬 Comments