Privacy

Privacy first,
not privacy theater

Blume is designed to help you understand how you use AI tools without building a permanent copy of your conversations in the cloud.

We believe you should know exactly what data is processed, why it's processed, and where it goes.

Your data starts local

Blume reads conversation transcripts from supported AI tools on your device and stores them locally in a SQLite database.

Your conversation history remains available on your machine, under your control.

When data leaves your device

Some Blume features require cloud-based AI processing.

When those features run, relevant portions of conversation data may be sent to AI models through our processing infrastructure in order to:

  • Generate titles and summaries
  • Extract metadata and insights
  • Analyze usage patterns
  • Power future intelligence features

We design these systems around data minimization:

  • Only the information required for a feature is transmitted
  • Requests are processed transiently
  • Conversation content is not stored after processing
  • Processing services are designed to be stateless whenever possible

When conversation data is sent off-device, we aim to make that behavior clear and predictable.

Cloud suggestion processing

If you explicitly choose Cloud for Improvements, Blume sends the conversation excerpts and project context needed for that analysis through our transient relay to Vercel AI Gateway and Google Gemini. The feature is designed so Blume does not retain that content or use it for training. Your suggestions and processing history remain on your device. Cloud processing will remain unavailable until the deployed provider route and these privacy controls are verified.

Before release, the relay must restrict requests to an approved Google model and require zero-data-retention and no-training routing. The release must not enable web search, grounding, file uploads, prompt caching, or provider fallback. Approved tools still run on your device, and only their bounded results are returned for the next model step.

Cloud is never selected automatically or used as a fallback. Your consent is stored on your device and can be withdrawn at any time by switching back to Connected plans. If the disclosure or data scope changes, Blume asks you to review it again before another Cloud run.

We don't build a cloud archive of your conversations

Blume is not designed to become a long-term repository of your chat history.

Conversation data may be processed to provide features, but we do not store copies of your conversations on Blume servers after processing is complete.

Your primary conversation archive remains on your own machine.

Usage and spend stay private

Usage and spending information is fetched directly from providers such as Anthropic, ChatGPT, and Cursor using credentials you control.

That information is displayed locally in Blume and is not transmitted to Blume servers for analytics or tracking.

We don't use your AI spending data to build user profiles, advertising systems, or training datasets.

Minimal product telemetry

Like most software products, Blume collects limited telemetry to understand crashes, performance issues, and product usage.

Telemetry is associated with a pseudonymous identifier rather than personal information and is designed to avoid collecting conversation content.

Our goal is simple: collect enough information to improve the product, and no more than necessary.

You control your data

You can:

  • Delete local Blume data
  • Reset your device identifier
  • Remove connected accounts
  • Stop using cloud-powered features
  • Remove Blume entirely from your system