[Feature] Please add commandcode.ai login provider

[agentsmithknowledgeisbliss-oss]

Area

AI providers and models

Problem

Can you please add command code login feature

Proposed direction

Can you please add command code login feature

Alternatives considered

No response

Additional context

No response

[luanweslley77]

Command Code now publishes a Provider API, so this does not need a custom login flow. It serves standard wires from one key at api.commandcode.ai/provider/v1:

• /chat/completions (OpenAI Chat Completions)
• /responses (OpenAI Responses)
• /messages (Anthropic Messages)
• /models (catalog; each entry carries supported_endpoints and context_length)

Auth is Authorization: Bearer <key> on every route (the key from the Command Code CLI works; /messages also accepts x-api-key). API access is included on every paid plan above Go, with per-model minimum plans: open models start lower, while Claude models require Pro or above. The catalog currently returns 80+ models split by wire: Claude models answer only on /messages; OpenAI and open models answer on /chat/completions, most also on /responses.

This maps onto the built-in provider architecture without new streaming code, since it is the same shape as the opencode entries, which already mix per-model api (openai-completions on the .../v1 root, anthropic-messages on the bare root). The work is a KnownProvider id (command-code), an API key env var, a catalog built from /provider/v1/models for ids, context windows, and wires, plus the usual login/display-name wiring. Reasoning efforts and advertised per-token rates are published in the model catalog documentation (they are not fields on the /models response), so thinkingLevelMap and cost can be seeded per model.

I have a key on a plan with API access and can validate the routes if that helps.