Skip to main content
OpenCode is an open source AI coding agent for the terminal, desktop, and IDEs. For AIOHub, configure a separate provider so AIOHub API keys stay isolated from your official OpenAI, Anthropic, or other provider credentials.

​
Prerequisites

  • OpenCode installed
  • An AIOHub sk- API key
  • A group assigned to the API key that can access the model you want to use
  • A positive account balance
Check that OpenCode is available: 
opencode --version
opencode --help

​
Config location

OpenCode supports global and project config. Use global config for your personal defaults and project config when you want to share model IDs with a repository.
Use caseFile
User-wide config~/.config/opencode/opencode.json
Current project configopencode.json
Project opencode.json overrides matching fields from global config. Do not commit real API keys. For shared project config, save credentials locally with /connect or reference an environment variable.

​
Configure Chat Completions models

OpenCode uses @ai-sdk/openai-compatible for OpenAI Chat Completions-compatible models. Set the Base URL to https://api.aiohub.org/v1, not a full endpoint URL.

​
1. Add the credential

In the OpenCode TUI, run: 
/connect
Choose Other, enter this provider ID: 
aiohub
Then enter your full AIOHub API key.

​
2. Add the provider config

Add this to opencode.json: 
{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "aiohub": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "AIOHub",
      "options": {
        "baseURL": "https://api.aiohub.org/v1"
      },
      "models": {
        "gpt-4o": {
          "name": "GPT-4o"
        }
      }
    }
  },
  "model": "aiohub/gpt-4o"
}
Replace gpt-4o with a Chat Completions model available to the group assigned to your API key. If you want a lightweight model for smaller tasks, add small_model: 
{
  "model": "aiohub/gpt-4o",
  "small_model": "aiohub/gpt-4o-mini"
}

​
Configure Responses models

If you want to use a model or workflow that requires OpenAI Responses, use @ai-sdk/openai. A separate provider ID keeps the Chat Completions and Responses paths clear. Put the API key in an environment variable: 
export AIOHUB_API_KEY="sk-your-api-key"
Then configure a Responses provider: 
{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "aiohub-responses": {
      "npm": "@ai-sdk/openai",
      "name": "AIOHub Responses",
      "options": {
        "baseURL": "https://api.aiohub.org/v1",
        "apiKey": "{env:AIOHUB_API_KEY}"
      },
      "models": {
        "gpt-5.3-codex": {
          "name": "GPT-5.3 Codex"
        }
      }
    }
  },
  "model": "aiohub-responses/gpt-5.3-codex"
}
You can keep multiple providers in one OpenCode config. Chat Completions and Responses use different request paths. If a model fails with endpoint or response-format errors, check that the npm package matches the model protocol.

​
Verify

Open OpenCode: 
opencode
Run this in the TUI: 
/models
Choose a model under aiohub/... or aiohub-responses/..., then send a short message. A normal reply means the setup is working. You can also verify API key and model visibility with HTTP: 
curl https://api.aiohub.org/v1/models \
  -H "Authorization: Bearer sk-your-api-key"

​
Troubleshooting

Confirm the provider ID entered through /connect exactly matches the provider key in opencode.json, for example aiohub. Restart OpenCode after editing config.
Set the Base URL to https://api.aiohub.org/v1. Do not set it to https://api.aiohub.org/v1/chat/completions or https://api.aiohub.org/v1/responses.
If you use /connect, confirm the credential is saved under the same provider ID. If you use an environment variable, run test -n "$AIOHUB_API_KEY" && echo "api key configured" before starting OpenCode.
Confirm the provider uses @ai-sdk/openai, not @ai-sdk/openai-compatible. If you only need basic chat, verify basic connectivity with a Chat Completions model first.

​
Official references