Configure Kimi K2.5, GLM 5, and MiniMax M2.5 as your OpenClaw inference provider — in 30 seconds. No Anthropic or OpenAI API key needed.
curl -fsSL https://github.com/getkimchi/kimchi/releases/latest/download/install.sh | bashOne command. Works on macOS and Linux.
kimchiKimchi auto-detects OpenClaw and configures it.
Your OpenClaw agent now runs on open-source models via Cast AI.
Curated for agentic workloads — large context windows, tool-use optimized, open-source.
Primary Agent
Reasoning, planning, code generation, image processing
Coding Subagent
Writing, refactoring, debugging code
Secondary Subagent
Code generation, debugging
| Feature | Kimchi | OpenRouter | Anthropic Direct | Ollama |
|---|---|---|---|---|
| Setup time | 30 seconds | 5–10 min | 5 min | 30+ min |
| Open-source models | Curated | Hundreds | Proprietary | Self-host |
| One API key | Local | |||
| Agent-optimized | Large ctx + tools | Generic | Expensive | HW-limited |
| Free tier | Free (your GPU) | |||
| Auto-config | Built-in | Manual JSON | Manual | Manual |
// ~/.openclaw/openclaw.json
{
"models": {
"providers": {
"kimchi": {
"baseUrl": "https://llm.kimchi.dev/openai/v1",
"apiKey": "${KIMCHI_API_KEY}",
"api": "openai-completions",
"models": [
{
"id": "kimchi/kimi-k2.5",
"name": "Kimi K2.5",
"reasoning": true,
"contextWindow": 262000,
"maxTokens": 32000
},
{
"id": "kimchi/glm-5-fp8",
"name": "GLM 5 FP8",
"reasoning": true,
"contextWindow": 202800,
"maxTokens": 32000
},
{
"id": "kimchi/minimax-m2.5",
"name": "MiniMax M2.5",
"reasoning": false,
"contextWindow": 196600,
"maxTokens": 32000
}
]
}
}
}
}Kimchi preserves your existing OpenClaw config. It only adds the kimchi provider — your other providers, channels, and agent settings stay untouched.
curl -fsSL https://github.com/getkimchi/kimchi/releases/latest/download/install.sh | bashkimchi and select OpenClaw