Bring Your Own LLM
Status
OpsWorker does not offer a generic, self-service "bring any LLM" backend today. Reasoning runs on Anthropic Claude models. There is no self-service support for OpenAI-compatible servers, GPT models, or self-hosted runtimes such as vLLM, TGI, or Ollama.
Supported Self-Service Options
| Option | What It Is |
|---|---|
| AWS Bedrock | The default managed backend. No setup required. |
| Custom Reasoning Model | Bring your own Anthropic API key, configured per cluster from the portal. Anthropic only. |
Other Model Hosting (On Request)
If you need a dedicated or private deployment, or a specific hosting arrangement that the options above don't cover, contact the OpsWorker team. These arrangements are scoped and set up with our team rather than configured self-service.
When we evaluate a model for investigation work, the main requirements are:
- Reasoning quality: investigation quality depends on the model's instruction-following and reasoning ability.
- Tool use: the model must support function calling / tool use for full investigation capabilities.
Next Steps
- LLM / AI Backend: overview of supported options
- Custom Reasoning Model: bring your own Anthropic API key per cluster