OpenCode

OpenCode is an open-source coding agent for the terminal (also available as a desktop app and IDE extensions) that you configure in opencode.json. Register MirAPI once as a provider and you can pick any MirAPI model per session, paying from your existing prepaid balance.

What you need

Element	Value
Base URL	`https://api.mirapi.ai/v1`
API key	Your MirAPI key, referenced as `{env:MIRAPI_API_KEY}`
Model	`deepseek-chat` under provider `mirapi`
Protocol	OpenAI-compatible (`/v1`)

If you do not have a key yet, create one in the Platform at https://platform.mirapi.ai. The key is shown only once, so copy it immediately.

Configure

Install OpenCode if it is not already installed:

curl -fsSL https://opencode.ai/install | bash

Other installers include npm install -g opencode-ai and brew install anomalyco/tap/opencode.

Add a provider block to opencode.json at your project root (or ~/.config/opencode/opencode.json for user-wide settings; JSONC is accepted):

{
  "provider": {
    "mirapi": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "MirAPI",
      "options": {
        "baseURL": "https://api.mirapi.ai/v1",
        "apiKey": "{env:MIRAPI_API_KEY}"
      },
      "models": {
        "deepseek-chat": {
          "name": "DeepSeek Chat"
        }
      }
    }
  }
}

Export the key in your shell:

export MIRAPI_API_KEY=sk-...

Set a default model

To skip the model picker on startup, add a top-level model key. OpenCode composes ids as provider_id/model_id, so a MirAPI model is prefixed with mirapi/:

{
  "model": "mirapi/deepseek-chat"
}

You can still override it per invocation with --model.

Verify with a message

opencode run "Reply with exactly: connected to MirAPI" --model mirapi/deepseek-chat

The reply should contain connected to MirAPI. In an interactive session, run /models and select mirapi/deepseek-chat.

How it works

OpenCode runs custom providers on the Vercel AI SDK. @ai-sdk/openai-compatible speaks the OpenAI Chat Completions protocol, so MirAPI receives POST /v1/chat/completions at https://api.mirapi.ai/v1. Responses stream as standard SSE (data: lines, ending with data: [DONE]), which the AI SDK parses for you.

Protocol variants

MirAPI serves the same models over three protocols. The provider above is OpenAI-compatible. If you prefer another SDK’s behavior, point the AI SDK’s @ai-sdk/anthropic provider at https://api.mirapi.ai (auth x-api-key) or @ai-sdk/google at https://api.mirapi.ai (auth x-goog-api-key). Whether a given model is available on a given protocol depends on its capability flags in the catalog.

Errors and billing

Status	Meaning	What to do
401	Key missing, invalid, or unknown	Check `{env:MIRAPI_API_KEY}` and the exported variable
403	Key valid but rejected: balance exhausted, model not whitelisted, quota reached, or IP not allowed	Resolve the cause, then retry
413	Request too large	Reduce the payload
429	Rate limited (no `Retry-After` header)	Retry with exponential backoff
500	Server error	Idempotent requests can be retried safely

Every error message ends with a request ID. Include it in any support ticket.

Chat usage is billed per token, with input, output, and cache read priced per 1M tokens. Reasoning tokens are billed as output. Prompt caching is automatic on OpenAI-compatible models; confirm hits via usage.prompt_tokens_details.cached_tokens. The balance is prepaid in USD and never expires. When it runs out, you get 403, and the same key works again immediately after a top-up. Reconcile spending in the Platform billing log or via GET /api/usage/token (account-level summary) and GET /api/log/token (per-request detail).

Tool-specific pitfalls

@ai-sdk/openai-compatible is the right npm package for OpenAI-compatible gateways; the plain openai package also works but keeps the SDK’s default behaviors.
The model list is manual. OpenCode does not auto-discover MirAPI models; add each one you want by its exact GET /v1/models name. The vendor/ prefix is Platform-only and never part of the request name.
The provider prefix is mirapi/. mirapi/deepseek-chat means the deepseek-chat model under the mirapi provider; it is not a model namespace.
403 with a valid key usually means the model is not whitelisted for that key or the balance is exhausted. Top up or adjust the whitelist, then retry.
Auth errors. Check that {env:MIRAPI_API_KEY} matches the exported variable name exactly.
Config changes need a restart. OpenCode reads opencode.json at startup, so restart it after editing the provider block.