Privacy Policy

Trident stores your workspace locally and connects directly to the AI providers you configure. It has no Trident account system, telemetry, analytics, automatic crash reporting, or Trident-operated server that receives your prompts. Local storage does not mean all processing stays on your computer.

Local data and credentials

Projects, documents, conversations, images, and embeddings are stored in an embedded database in Trident’s user data directory, with project documents and images also saved as ordinary files under ~/Trident/. Workspace content is not encrypted by Trident.

Provider credentials are encrypted locally using your operating system’s secure storage. Trident uses API keys or derived authentication tokens to authenticate with your configured providers and their identity services. Regions, endpoints, and model identifiers are stored as ordinary settings.

AI provider requests

Chat and image generation send your prompts, conversation context, and any documents or images included in the request directly to the selected provider. Connections include Anthropic, OpenAI, and Google Gemini APIs, and Amazon Bedrock, Google Vertex AI, and Azure OpenAI through your cloud account. Connections may also contact authentication and model-discovery endpoints. Processing and retention are governed by your provider account and its terms; this traffic does not pass through Eastechs infrastructure.

OpenAI semantic indexing and search

With an OpenAI key configured and semantic indexing enabled for a project, saving or editing documents sends their text chunks to OpenAI for embeddings. Image indexing sends image names and generation prompts. This happens independently of the chat provider selected in either panel.

Disabling semantic indexing stops new automatic document and image indexing; requests already in progress may finish. Semantic searches separately send the search query to OpenAI, including searches over an existing index.

Background requests and remote content

• GitHub Releases and its download hosts: update checks on launch and every four hours, plus update downloads. Metadata requests include a locally generated, persistent installation identifier (x-user-staging-id) used for staged rollouts.
• raw.githubusercontent.com: refreshed model pricing data.
• models.dev: provider logos in the model picker.

These services receive ordinary network metadata such as your IP address and request headers, but no workspace content or provider credentials. Remote images embedded in documents or model responses may contact their hosts. External links open in your default browser.

This website and feedback

This website loads fonts from Google Fonts. Its hosting provider and Google receive ordinary web request information, such as your IP address and request headers. The site does not include analytics or advertising scripts. Downloads, sponsorships, issues, and discussions link to GitHub, whose privacy policy applies there.

If you contact us by email or post feedback, we receive the information you choose to provide. GitHub issues and discussions are public. Please report security vulnerabilities through private vulnerability reporting or security@eastechs.com.

Changes and contact

We will update this policy when Trident’s data handling changes. Questions about this policy? Email hello@eastechs.com.