Using Corion AI with Kimi Code

Kimi Code CLI — Moonshot's terminal coding agent — can import all of Corion's models as a provider in one command, straight from our registry file.

One-line setup (recommended)

Get an API key from Dashboard → API keys, then run:

kimi provider add https://corion.ai/kimi-registry/api.json --api-key YOUR_KEY

That's all. The command creates the Corion provider and imports every model Corion offers — kimi-k3 (1M context), deepseek-v4-pro-0813 and deepseek-v4-flash-0731 — as select-ready model aliases.

Confirm your key sees the models before driving code changes:

curl https://api.corion.ai/v1/models -H "Authorization: Bearer YOUR_KEY"

Then:

Pick a model for one launch: kimi -m corion/kimi-k3
Or choose interactively in the TUI with /model

Run a read-only task first ("summarize this repository") and confirm the reply and tool calls behave before you let Kimi Code edit files.

Why this way

The same registry also powers the interactive flow if you prefer clicking around: /provider → Add New Platform → Custom registry, then paste the URL and your key.

And because Corion's provider metadata stays tied to the registry URL, later launches of Kimi Code auto-refresh models and metadata from the registry — new models, context changes, and price updates show up without touching your config again.

The registry is public — no key is needed to fetch it. The --api-key value is stored as the credential Corion requests are billed against.

The registry file

Corion serves a models.dev-schema registry at a stable URL: https://corion.ai/kimi-registry/api.json. Fetch it directly to inspect exactly what Kimi Code imports:

{
  "corion": {
    "id": "corion",
    "name": "Corion AI",
    "type": "openai",
    "api": "https://api.corion.ai/v1",
    "doc": "https://corion.ai/docs",
    "models": {
      "kimi-k3": {
        "id": "kimi-k3",
        "name": "Kimi K3",
        "tool_call": true,
        "reasoning": true,
        "interleaved": { "field": "reasoning_content" },
        "attachment": true,
        "modalities": { "input": ["text", "image"], "output": ["text"] },
        "limit": { "context": 1048576, "output": 131072 },
        "cost": { "input": 2.5, "output": 12.5, "cache_read": 0.25 }
      },
      "deepseek-v4-pro-0813": {
        "id": "deepseek-v4-pro-0813",
        "name": "DeepSeek V4 Pro 0813",
        "tool_call": true,
        "reasoning": true,
        "interleaved": { "field": "reasoning_content" },
        "attachment": false,
        "modalities": { "input": ["text"], "output": ["text"] },
        "limit": { "context": 1048576, "output": 384000 },
        "cost": { "input": 0.435, "output": 0.87, "cache_read": 0.0435 }
      },
      "deepseek-v4-flash-0731": {
        "id": "deepseek-v4-flash-0731",
        "name": "DeepSeek V4 Flash 0731",
        "tool_call": true,
        "reasoning": true,
        "interleaved": { "field": "reasoning_content" },
        "attachment": false,
        "modalities": { "input": ["text"], "output": ["text"] },
        "limit": { "context": 1048576, "output": 65536 },
        "cost": { "input": 0.11, "output": 0.25, "cache_read": 0.011 }
      }
    }
  }
}

limit.context, limit.output, and cost per model are read by Kimi Code as authoritative metadata; cost values are the Corion rate card in USD per million tokens.

Model aliases combine the provider ID and model ID: corion/kimi-k3, corion/deepseek-v4-pro-0813, corion/deepseek-v4-flash-0731.

Troubleshooting

401 Unauthorized — check the key: it must be the complete key shown once at creation. Create a fresh one in the dashboard if in doubt.
Model not listed after refresh — re-run the kimi provider add command, or remove and re-add the provider (kimi provider remove corion) to force a fresh import.
New Corion models not appearing — Kimi Code refreshes providers from the registry URL on startup; restart the CLI.

Sources: Providers and models — custom registry, kimi provider reference.