Provider API
Provider API
Build with Command Code's models anywhere. Now available via API.
//how it works
Three steps to your first call
Follow these steps to integrate the Command Code Provider API.
Get an API key
Generate a key in Command Code Studio. Same key as the CLI. Provider plan or higher.
Point your client at the URL
Set the base URL, select an endpoint, and pass your key as a bearer token.
Calls draw from your plan
Every request bills against your Command Code credits at API rates. It all shows up in Studio.
//endpoints
Three endpoints
Provider API has two endpoints for OpenAI and Anthropic formats. You can also fetch the available list of models via the models endpoint.
| Endpoint | Method | Format |
|---|---|---|
| https://api.commandcode.ai/provider/v1/chat/completions | POST | OpenAI Chat Completions |
| https://api.commandcode.ai/provider/v1/messages | POST | Anthropic Messages |
| https://api.commandcode.ai/provider/v1/models | GET | Models list |
//models
The best models, from the labs that build them
//pricing
Start with the Provider plan
$15/month includes Provider API access, $15 in credits, then pay as you go. See the plans for tiers and limits.
Provider plan
$15/month. Includes Provider API access, $15 in credits, then pay as you go.
API rates, no markup
Calls spend plan credits at underlying API rates. Top-ups roll over. They never expire.
Tracked in Studio
Every request, token, and cost. On your Studio usage timeline.
//cli vs api
When to use the Provider API
A direct API call gets you the model. The CLI gets you the whole harness. Pick the one that fits the job. Read the Provider API launch post.
Prefer the CLI for coding
For real coding work, stick to Command Code CLI. It has the best coding harness with taste, context, and tool execution tuned per model.
Automating? Go headless
Scripts, CI, automation? Run the CLI headless. One query, straight to stdout, then it exits. No agent to rebuild around an API.
cmd -p "review this diff"Reach for the Provider API
When the job lives outside the terminal, reach for the API. Your plan and credits, in any OpenAI or Anthropic client.
Developers love it. Founders, too.
“Command Code learns my taste. After a week, it stopped making the mistakes I kept fixing in other agents. The diffs feel like a senior engineer who already read the codebase.”
“Command Code is the first agent where I trust open models in production. The harness is so solid I had to double check I was still on DeepSeek Flash. Shipped multiple CLIs and a full Redwood app for $2.”
//faq
Questions, answered.
Still curious? Read the Provider API docs. Or email support@commandcode.ai.
Provider plan or higher. Accounts on the Go plan get a 403 upgrade_required on every request and need to upgrade before they can call the API.
No. Use the same Command Code API key you already have for the CLI. Create an API key here.
Anthropic, OpenAI, and open-source models. Each endpoint accepts only its native wire shape: /chat/completions serves OpenAI and open-source models, and /messages serves Anthropic models. Sending a Claude model to /chat/completions returns a 400 directing you to /messages, and vice versa.
No. The endpoint returns one completion per request. n > 1 is rejected at the schema layer. Send parallel requests if you want multiple samples.
Text and image. Audio, file, and document parts are rejected by the schema. If the chosen model doesn't support vision, the upstream provider returns its own error.