How we protect your code and data

Tenant isolation

• Every customer is a separate tenant. Data is isolated at the database layer using PostgreSQL Row-Level Security (RLS), enabled and forced on every tenant-scoped table — so a query can only ever see its own tenant's rows, even in the event of an application bug.
• Each request resolves your tenant from a signed identity claim and binds it for the life of the database transaction. There is no cross-tenant query path.
• Each project's working files live under a per-tenant, per-project storage prefix in isolated cloud storage.

Your keys, your models — we don't train on your code

• You bring your own LLM provider keys. They're stored as per-tenant secrets in Google Secret Manager (one entry per tenant per provider) and used to call the provider directly on your behalf.
• We do not train any models on your code, prompts, or data. Inference runs against the providers you configure under your own account and their data-use terms.

Network & encryption

• All traffic is encrypted in transit over TLS. Data is encrypted at rest by the underlying cloud platform.
• The database is reachable only over a private network — it has no public IP. Application services reach it through a private VPC connection.
• Continuous-integration deploys authenticate via Workload Identity Federation (short-lived, keyless credentials) — there are no long-lived cloud keys in CI.

Governance over autonomous agents

An autonomous agent touching your codebase should be governed, not unleashed. BuilderBots gives you the controls:

• Human approval gates — in Supervised mode, changes wait for an explicit human approve/reject before they're merged.
• Stop & abort — you can abort any task (and its sub-tasks) at any time; a kill switch halts work at the next safe checkpoint.
• Hard budget caps — per-task, per-project and daily dollar caps are enforced mid-flight, so spend can't run away. New workspaces start with safe caps you can tune.
• Quality gates — every change runs through a multi-stage pipeline (tests, verification, and adversarial code review) before it can merge.
• Honest status — agents may not claim a task is running, complete, or shipped unless the system's own records confirm it.

Authentication & access

• Authentication is handled by Google Identity Platform (Firebase Auth) with HttpOnly, Secure session cookies. Sessions are checked for revocation on each request, so disabling an account or signing out takes effect immediately.
• Multi-factor authentication and enterprise SSO (SAML/OIDC) with SCIM provisioning are on our near-term roadmap — talk to us if you need them for a rollout.

Data residency, export & deletion

• Production data is hosted on Google Cloud in the United States (us-central1).
• You can delete a project's workspace from the app at any time. To export your data or delete your account and all associated data, email security@builderbots.ai — see our Privacy Policy.

Sub-processors

We rely on a small set of infrastructure providers to operate the service:

• Google Cloud Platform — compute (Cloud Run), database (Cloud SQL for PostgreSQL), object storage, and Secret Manager.
• Google Identity Platform (Firebase Auth) — authentication.
• Your configured LLM providers (e.g. Anthropic, OpenAI, Google) — model inference, called with your own keys under your account.

Compliance

BuilderBots is an early-stage product. The controls above are live today; formal third-party attestations (e.g. SOC 2) and a signable DPA are on our roadmap. If your procurement process needs specific documentation, reach out at security@builderbots.ai and we'll work with you.