Free, open-source, privacy-first AI usage tracking browser extension.
Effective date: September 17, 2026 · Last updated: September 17, 2026
Treco is a free, open-source, non-commercial browser extension, developed and maintained by an individual developer — not a company, and not operated through any registered business entity. This Privacy Policy explains what Treco does, what data it processes, and what it never collects. It describes the extension exactly as implemented today, not an idealized version of it.
Treco is a browser extension that:
Treco is provided free of charge. There is no paid subscription or tier, no advertising, no account system, no analytics or telemetry, and no server-side storage of any kind — Treco has no server or backend at all.
For each detected interaction, Treco’s background service worker constructs and stores a small usage record containing:
Aggregated historical statistics (totals, trends, streaks) shown on the dashboard are computed locally, on demand, from these stored records — they are not a separately collected dataset.
Treco also stores whatever you explicitly configure: which providers are enabled, your unit preference, theme, any optional usage goals you set, and an optional local electricity-grid carbon-intensity override.
We want to be precise here rather than reassuring in a way that isn’t technically accurate. To estimate token counts, Treco’s content script needs to read the rendered text on the page of a supported AI site while you’re using it — this can include the visible text of your messages and the AI’s replies, because that is the only way to measure roughly how much text was involved.
That text is read transiently, in your browser, purely to calculate a character/token count. It is never written to storage, never logged, and never transmitted anywhere — only the resulting numbers (token counts, not the text itself) leave that calculation. As a technical safeguard, every record written to storage is validated against a strict allow-list of fields before it’s saved; any attempt to store a long text field or an unrecognized field is rejected automatically, so conversation content cannot end up in Treco’s storage even by accident.
All usage records, settings and goals are stored using your browser’s local extension storage
(chrome.storage.local or the equivalent in your browser), on your own device. Treco does not use any cross-device
sync storage, and there is no remote database or server that your data is uploaded to — none exists. If you
uninstall the extension, this locally stored data is removed by the browser along with it.
Treco does not currently include any analytics, telemetry, crash reporting, or error-logging service of any kind — none is built into the extension, and none runs in the background. If that ever changed in a future version, it would be off by default, clearly disclosed in an update to this policy, and controllable from Settings.
Treco’s content script runs on the AI provider pages you visit so it can observe usage there, but Treco does not make network requests to those providers, to any Treco-operated infrastructure (none exists), or to any other third party. Treco does not have access to your account, billing, or conversation data on any AI provider’s own systems — only to what is visibly rendered on the page while you’re using it, as described above.
The providers currently supported are listed in the extension’s Settings → Providers page; the extension only requests browser permission to run on those specific domains, and only observes a provider’s page if you’ve left that provider enabled.
The energy, CO2e and water figures Treco shows are estimates, not direct measurements. No AI provider currently publishes per-request telemetry precise enough to measure these values exactly, so Treco derives them from published research, provider-disclosed figures where they exist, and documented model-specific or fallback-tier coefficients otherwise. Every figure in the UI is labeled with a confidence/method indicator reflecting how it was derived.
These estimates can vary meaningfully depending on the specific model, the provider’s underlying infrastructure and hardware, real-world workload and load conditions, the geography and electricity mix of the data center actually used, and the calculation methodology itself. See Methodology for the full calculation and its sources and assumptions.
Treco is built around minimizing what it collects in the first place: it requests only the browser permissions it needs (local storage), only requests host access to the specific AI provider domains it supports, and validates every record against a strict schema before storing it so that unexpected or oversized data — such as accidental conversation content — is rejected rather than saved. Because your data stays on your device, its security also depends on the security of your own browser profile and device.
Treco is not directed at children and is not intended for use by anyone under the age of 18. Treco does not knowingly collect personal information from children, and given it has no account system, no personal information is knowingly collected from anyone.
If this policy changes, the date above will be updated and, for material changes, the change will be noted inside the extension (for example, on the Settings or About page) with the next update. We encourage checking back periodically.
Treco’s source code is public at github.com/medaharrat/treco. It is licensed under the MIT License.
Treco is operated by an individual developer, not through a registered company or commercial entity. There is no corporate address to disclose here because none exists — this policy is issued directly by the developer, based in Hungary.
Questions about this policy or Treco’s data practices can be raised via the project’s public GitHub repository — see Contact for the exact link and what to include in a report.