Built to be honest.

Your data belongs to you

This isn't a terms-of-service footnote. It's the design principle.

• Your conversations and files belong to you, always.
• We do not use your data to train AI models.
• Your data is stored in Supabase (PostgreSQL), solely to provide the continuity features you choose to use. You can review Supabase's privacy policy.
• You can export or delete your data at any time.

Many AI tools quietly retain or reuse user content. Continio is built around the principle that what you share stays yours.

How Continio uses AI models

Continio routes your conversations to trusted third-party AI providers to generate responses. These currently include OpenAI and Anthropic. You can also choose a specific model manually.

When you send a message, only what is needed to generate that response is shared. We do not send your full memory or file library automatically.

• Relevant memory context is selected and included, not your entire history.
• Workspace files are only sent to the AI when you explicitly attach them to a message. They are never injected silently.
• Per-thread file uploads are sent only within that conversation.
• We do not use your conversations to train AI models.

We do not build the models. We design how they are used, and we are deliberate about that distinction.

You can review the privacy policies of our model providers: OpenAI  ·  Anthropic

AI and the environment

AI products use real compute, which uses real energy. We are not going to pretend otherwise.

What we can say honestly is that Continio is designed to use less compute per useful output than the alternative, and we make deliberate technical choices that reduce energy use per message.

• Continuity reduces redundancy. The biggest source of wasted AI compute is re-explaining context. Every time a user starts a new chat and re-introduces themselves, that is compute spent re-processing information the system should already know. Continio eliminates this. Fewer redundant tokens means less energy per useful response.
• Prompt caching. Continio caches the system prompt sent to the model on every message. A cached prompt costs a fraction of the energy of a full inference - saving approximately 90% of the compute cost of the system prompt across repeated calls. In March 2026 the system prompt was also compressed by 50%, permanently halving this cost.
• Ranked memory injection. Continio stores a growing picture of who you are. Rather than injecting all of it into every message, it scores and selects only the most relevant items for each request. A simple question might use 4 memory items. A reflective conversation might use 12. The rest stays quiet, reducing tokens per message as your memory grows.
• Model routing. Not every message needs the largest model. Simple questions go to lightweight models. Complex or personal conversations get the full model. This is primarily a quality decision, but lighter models also use meaningfully less compute.
• Background throttling. Enrichment tasks pause when you are actively chatting. They run during idle periods, avoiding unnecessary concurrent compute load.

Continio runs on Vercel, Railway, and Supabase. AI inference runs on Anthropic and OpenAI infrastructure, both of which have published net-zero and renewable energy commitments.

We are not carbon neutral and we have not published an emissions figure. We do not have the data to do so accurately, and publishing a number we cannot verify would be worse than saying nothing. What we commit to is this: as the product grows, efficiency improvements come before feature additions.

Read more in the Knowledge Base → AI and the environment.

Security: an honest account

Continio is built by a solo founder using AI-assisted development. That context matters when assessing security, and I'd rather be straight about it than leave you to guess.

What is in place: Authentication is handled by Supabase with JWT verification on every request. Every database query is filtered by user ID. Row-level security is enforced on all database tables and verified on every deploy. Payments are handled entirely by Stripe - card details never touch Continio's infrastructure. Rate limiting is active. Auth tokens are never written to localStorage.

The honest gap: The codebase has not yet had an independent security review by a professional. I run regular checks against a security checklist and fix what I find, but I cannot replicate the pattern recognition of someone who has spent years finding non-obvious vulnerabilities. An independent review is committed before Continio moves from early access to general availability. I will publish a note here when that happens.

For most everyday use - thinking through problems, writing, planning - the current posture is appropriate. If you are working with genuinely sensitive professional or legal material, I recommend keeping that in a separately secured system until the independent review is complete.

What we don't do

Sometimes the clearest signal of intent is what a product refuses to build.

Your information exists to support you, not to profile you.

Safety and responsible use

AI systems can be misused. Continio is designed with this in mind.

• We aim to be transparent when memory or past context is used in a response.
• We avoid hidden decision-making.
• We design to support human judgment rather than replace it.
• We continuously review design choices to ensure they align with responsible use.

Your responsibility as a user

Continio is a tool to support your thinking. You remain responsible for the decisions you make with the information it provides.

• Verify important decisions independently.
• Avoid sharing sensitive personal data unless necessary.
• Treat AI as a support tool, not a sole authority.

Our commitments

What we believe

Trust and data policy tell you what we do. The principles page tells you why. Six things that guide every product decision we make, written plainly and held to publicly.