Commercial licensing and deployment
For engineering teams evaluating Robota: review licensing, deployment options and the data flows to verify before use.
Commercial licensing
Robota is available under AGPL-3.0 or a commercial license for uses that cannot meet AGPL obligations. Open a GitHub Discussion to ask about licensing; do not include credentials or sensitive security details.
Deployment and data flow
Data Handling
The CLI runs on your machine and the SDK runs inside your host application. Data flow depends on the provider and tools you configure — review those connections before deployment.
| Data | What to verify |
|---|---|
| Prompts and supplied context | Sent to your configured model endpoint; a hosted endpoint receives the data you send. |
| Provider credentials | Supplied by your environment or host configuration; verify storage and access rules. |
| Session history | The CLI uses configured local state. SDK persistence is optional and owned by the host. |
| Tool results and integrations | A tool result can become model context; web or MCP tools may contact other services. |
On-Premises Deployment
You can connect to a local OpenAI-compatible model server. Restricted-network use also requires local packages/models and review of any enabled network tools:
- • Ollama — run a downloaded model on your network
- • LM Studio — run a local OpenAI-compatible server
- • Compatible gateway — point baseURL to a server you operate
import { OpenAIProvider } from '@robota-sdk/agent-provider/openai';
const provider = new OpenAIProvider({
apiKey: 'local',
baseURL: 'http://your-internal-gateway/v1',
defaultModel: 'your-loaded-model-id',
});
Public source
Inspect the repository and dependency choices before deployment.
Configurable model endpoint
Choose a hosted provider or a local compatible server.
Host-owned SDK state
The SDK host decides whether and where to persist sessions.
Explicit tool permissions
Review permission modes and enabled integrations for the deployment.
FAQ
Where do prompts and code go?
Model requests go to the endpoint you configure. A hosted provider receives the prompts and context sent to it; a local model can keep inference on your own server. Enabled web, MCP or other tools can make separate network requests.
Can we use a corporate proxy?
The model endpoint and each enabled integration must fit your network policy. Verify proxy behavior in your chosen configuration before rollout.
Can this run in a restricted network?
A local model server and internally available npm packages can support a restricted setup. Disable or configure network-capable tools according to your policy.
Is there a commercial license option?
Yes. The public source is available under AGPL-3.0, with a commercial license option for proprietary or closed-source use that cannot meet AGPL obligations. Review the licensing terms before choosing a path.