Transparency

Our Commitment

Eco is built on radical transparency. We publish exactly how the system works because privacy claims are only meaningful when they can be verified. Every line of code is open source under the AGPL-3.0 license — anyone can inspect, audit, and confirm our architecture enforces what we promise.

How Eco Works

Eco web v1.0 runs the AI model on your device, inside your browser. Your conversations are not sent to Eco servers for inference.

• On-device inference — your query runs entirely in your browser. Your prompts and responses do not leave your machine to be processed.
• Local storage — conversations you keep are stored in your browser on your own device, and you can clear them at any time.
• Model downloads — to run a model locally, your browser downloads model files. Those requests carry model identifiers, not your conversation content.

Your conversations stay on your device. The AI runs in your browser, so we never receive your prompts or responses to generate an answer.

Local model setup still downloads model artifacts through Eco's reviewed proxy and upstream model storage. Those requests are for model files only; they are not prompt, file, conversation, or response egress.

Web lookups (optional, on by default). For factual questions, Eco can check Wikipedia and Wikidata so it cites a real source instead of guessing, and for weather questions it can check Open-Meteo. When this is on, your browser fetches the search terms from your question directly from Wikimedia — and, for weather, sends the city you ask about (never your device location) to Open-Meteo. Eco's servers never see or store any of this, and your full conversation is never sent; only those search terms or the named city reach each provider, subject to its own privacy policy, much like visiting the site yourself. Turn it off in Settings → Eco to keep every request fully on your device; weather lookups stop too.

On-Device Inference

What We Can See

The Eco API gateway has access to the following operational data:

• Account email — if you register an account (optional for local use)
• Subscription and billing status — whether you support Eco with a paid subscription, handled through Stripe
• Operational metadata — request timestamps, IP addresses, and HTTP metadata used for security, rate limiting, and abuse prevention

What Eco Operators Cannot See

The following data is designed to stay out of Eco operator view:

• Your conversation content — the AI model runs in your browser, so your prompts and responses are not sent to us. We have no mechanism to access them.
• Your local conversation history — stored in browser storage on your device. We don't have access to your local browser storage.
• What models you use or how often — local inference is invisible to us. We can't see what runs on your hardware.

The Code

Eco is fully open source under the GNU Affero General Public License v3.0 (AGPL-3.0). Every line of code is auditable. Our privacy architecture can be independently confirmed — you never have to take our word for it.

If you find something that contradicts our launch claims, please tell us.

Infrastructure

In the spirit of transparency, here is where your data lives and who processes it:

• Web app — hosted on Vercel. The AI model runs in your browser; chat inference is not processed on the server.
• API gateway — hosted on Fly.io. Handles authentication and billing only; it does not process your conversations.
• Database — PostgreSQL on Neon (accounts and billing records — never conversation content)

Contact

Questions about our transparency practices or privacy concerns? Contact us at transparency@eco.network. We take every report seriously and will respond within 48 hours.