We don’t want your data in the first place.

1. The short version

Lumium is local-first. The console and runtime store your data on your machine. We do not collect prompts, completions, tool inputs, tool outputs, file contents, or anything you type into an agent. We do not phone home. We do not train models on you.

2. What we collect

When you visit lumium.org we serve static pages with no third-party trackers. Standard server-side request logs (IP, user-agent, URL, timestamp) are retained for 30 days for security and abuse prevention, then deleted.

When you fill out the contact form, we receive the fields you submit and your IP. We use those to reply to you and nothing else. The form does not write to a CRM or third-party processor in this version.

3. What the console collects

Nothing. There is no telemetry, no crash reporting, and no usage analytics.

4. Provider data

When an agent calls a model provider (OpenAI, Anthropic, Gemini, Cohere, Ollama, etc.), the request goes directly from your machine to that provider using your API key. Lumium does not proxy, mirror, or store provider traffic. Whatever your provider does with your prompts is governed by their terms.

5. Subprocessors

The marketing site is hosted on a CDN. Email replies use a transactional email vendor. We will publish the subprocessor list before v1.0 ships.

6. Your rights

If you are in the EU, UK, California, or another jurisdiction with applicable data rights, you can email privacy@lumium.org to access, correct, or delete any data we hold about you.

7. Changes

If we change this policy in a material way, we will publish the change in the changelog and update the date at the top.