Transparency
How your data flows
Three modes are supported. Each one routes your code differently. Local mode keeps AI inference on your machine; cloud BYOK sends selected code to the provider you chose.
Local mode (Ollama / LM Studio)
Your codeLocal AI modelDecoder UI
Cloud BYOK mode (OpenAI / Anthropic / Gemini / OpenRouter)
Your codeDecoder backendSelected AI providerDecoder UI
GitHub import mode
GitHubDecoder importDecoder UI
You authorize GitHub. Decoder imports the metadata and file contents of the repositories you select, under the scopes you granted. You can delete imported project data at any time.
Before you upload
- Do not upload secrets, tokens or `.env` files.
- Remove `.env`, credentials and customer data from any archive.
- Review the AI provider's terms before sending real code.
- Start with a public demo repo or non-sensitive code.