Integrations
OpenCode is an open-source coding agent for the terminal. Add DEVUP AI to it as a custom OpenAI-compatible provider.
Install OpenCode globally using npm, then verify the installation:
npm install -g opencode-ai
opencode --versionSet your DEVUP AI API key as an environment variable in your shell:
export DEVUP_API_KEY="your-key"Configure DEVUP AI in opencode.json:
{
"$schema": "https://opencode.ai/config.json",
"provider": {
"devupai": {
"npm": "@ai-sdk/openai-compatible",
"name": "DEVUP AI",
"options": {
"baseURL": "https://api.devupai.com/v1",
"apiKey": "{env:DEVUP_API_KEY}"
},
"models": {
"deepseek-ai/DeepSeek-V4-Pro": {
"name": "DeepSeek V4 Pro",
"limit": { "context": 1048576, "output": 8192 }
}
}
}
},
"model": "devupai/deepseek-ai/DeepSeek-V4-Pro"
}| Field | Meaning |
|---|---|
| devupai | provider ID |
| npm | @ai-sdk/openai-compatible |
| baseURL | https://api.devupai.com/v1 |
| apiKey | {env:DEVUP_API_KEY} reads the environment variable |
| model key | deepseek-ai/DeepSeek-V4-Pro, the exact catalog ID |
| limit.context / limit.output | the limits OpenCode uses for this model: context is the model's context_window, and output at most its max_output_tokens, both in GET /v1/models |
| model | devupai/<model-id>, the default model |
Place the file in your project root, or in ~/.config/opencode/opencode.json (Windows: %USERPROFILE%\.config\opencode\opencode.json) for all projects.
Run opencode models devupai to verify that OpenCode recognizes the configured provider:
opencode models devupaiThis command lists devupai/deepseek-ai/DeepSeek-V4-Pro as an active model.
Start the interactive interface by running opencode. OpenCode opens with DEVUP AI selected:
Enter your task prompt in the interactive prompt bar:
Create hello.py that prints "Salam DEVUP", then run it.OpenCode executes tools to write the file, executes the shell command, and displays context usage:
Add more models under models using their exact catalog IDs. For agent tasks, choose models whose catalog entry supports tool calling. Browse full details in Tool Calling and Models.