Security
In one paragraph. Your account gets its own isolated environment for every build and preview. Your secrets and connected credentials are encrypted at rest and never shown back in full. Anything an agent does to a production system goes through a brokered, audited path with the smallest privilege that does the job. Hosted apps run on hardened, patched infrastructure behind TLS. We do not train AI models on your code or data. And when something goes wrong, the status page says so.
- Isolation by design
- Secrets and credentials
- What agents can and cannot do
- Production access
- Hosted apps
- Your account
- Data handling and AI
- Operations, monitoring and incidents
- Reporting a vulnerability
- Compliance
1. Isolation by design
Every account's builds, previews and agent sessions run inside a sandbox dedicated to that account, on isolated compute with its own file system and network namespace. An agent working on your app cannot see another customer's code, environment or previews. Previews are served from a separate domain reserved for customer content, so they never share cookies or browser trust with the Nanza app itself.
Inside your sandbox the agents have what they need to build and test your app. Outside it, nothing: the control plane, other customers, and your production systems are all separate boundaries that the agent can only reach through the brokered paths described below.
2. Secrets and credentials
- Connected credentials (your AI provider, GitHub, cloud accounts, domain registrars, messaging providers) are encrypted at rest with keys that live outside the database.
- The secrets your app needs (API keys, service credentials) are stored in your app's own encrypted credentials file, committed to your repository in encrypted form. The platform shows key names, never values, and agents request a secret from you rather than inventing or reading it.
- Nanza never sees your card number: payments are entered directly with Stripe.
- Everything travels over TLS: nanza.com, previews, hosted apps, and every call to a provider.
3. What agents can and cannot do
Agents are given a task, a sandbox, and a set of tools. The prompts that govern them forbid actions outside the project they are working on, forbid privilege escalation on shared machines, and require them to ask you for anything they lack rather than guess. Those rules are enforced in the environment, not just requested: the sandbox is the boundary, and on shared machines agents run as a user with no administrative rights.
Every agent action is recorded in the task's execution log, which you can read in the app. Changes to your app arrive as reviewable work on your board and reach production only when you approve them or when your workflow's topology deliberately allows it.
4. Production access
Reading data from your app's production database, or changing it, is the most sensitive thing the platform can do on your behalf, so it goes through one brokered path:
- Reads use a dedicated read-only database role, with statement time limits and output caps.
- Writes require a separate permission on your account, run inside your app's own container under your app's own role, and take a restore point first.
- Every call, read or write, is recorded with who asked, what ran, and what came back.
Nanza staff do not read your production data in the ordinary course of business. Access for support or incident response is deliberate, limited to what the situation needs, and recorded.
5. Hosted apps
- Published apps run on dedicated infrastructure in the United States, one server per app, with the operating system kept patched.
- Traffic terminates at a proxy with an automatically issued and renewed TLS certificate, on a Nanza address or your own domain.
- Each app has its own database with its own credentials; databases are not reachable from the public internet.
- Apps ship with error tracking, structured logs, metrics and traces reporting to our monitoring, so a failing app is noticed by the platform, not just by your users.
- Container images are built from your repository on isolated build infrastructure and stored in a private registry.
6. Your account
- Passwords are stored as salted hashes and never logged. Sign in with Google is available if you prefer not to keep a password with us.
- Sessions are server-side and can be ended from any device by signing out; a session that is no longer valid is refused on the next request.
- Team members get exactly the permissions their role grants, and a membership can be limited to specific projects.
- API tokens are scoped to a subset of your permissions, stored only as a hash, and revocable at any time.
- Signup is protected against automated abuse.
7. Data handling and AI
Your code and data are yours, and we do not use them to train AI models. The AI providers that run the agents process your prompts and code under commercial terms that do not permit training on that data. The Privacy Policy lists every provider we use, what each receives, and how long we keep things.
8. Operations, monitoring and incidents
Nanza runs on Nanza: the platform monitors itself with the same tooling it gives your apps, watches every hosted app and server from the outside, and opens an incident for a human when something needs one. Deploys are automated and verified before traffic moves. Platform-wide problems are declared on the status page, with updates until they are resolved, and customers affected by a security incident involving their data are told without undue delay.
9. Reporting a vulnerability
If you believe you have found a security issue in Nanza, in a preview, or in a hosted app's platform layer, email hello@nanza.com with "Security report" in the subject. Please include what you found, how to reproduce it, and how we can reach you. We will acknowledge your report within two business days, keep you informed while we work on it, and credit you if you would like. We ask that you give us a reasonable time to fix an issue before disclosing it, avoid accessing other customers' data, and do not run tests that degrade the service for others.
10. Compliance
Nanza is a young company. We have not yet completed a third-party audit such as SOC 2, and we would rather say so than imply otherwise. The practices on this page are how we operate today; as we grow, we will publish audit reports and certifications here as we obtain them. If your organization needs a security questionnaire completed or a data processing agreement, email hello@nanza.com.