Isolated by default. Connected only by you.

Aria runs a whole company of autonomous agents on your behalf. Because those agents take real actions — reading code, opening pull requests, deploying — security can't be an afterthought. It's the foundation the platform is built on.

The guiding ideas are simple: isolate everything, grant the least access necessary, and never touch what you haven't connected. Below is how that plays out in practice.

Each agent operates inside its own isolated sandbox — a contained execution environment with its own filesystem and process boundary. One workspace cannot see or reach into another, and a single agent's blast radius is confined to the sandbox it runs in.

• Tenant isolation.Your company's agents and data are separated from every other customer's at the infrastructure level.
• Bounded execution.Agents run within defined resource and permission limits — they can't escape the sandbox or reach hosts you haven't connected.
• Ephemeral by design.Sandboxes are provisioned for the work at hand and torn down when it's done.

Aria has no standing access to your systems. Agents can only touch a repository, service, or resource after you explicitly connect it and approve the scope. Nothing is connected by default.

• You hold the keys. You choose what to connect, you approve the permissions, and you can revoke access at any time.
• Default-deny.If a resource isn't connected, an agent's attempt to reach it is denied — and logged.
• Full audit trail. Agent actions are attributed and recorded so you can review what was done, when, and why.

When you connect a code host such as GitHub, Aria requests the minimum scope needed to do the work you've asked for, on the repositories you select — not your whole account.

• Repository-scoped.Access is limited to the specific repositories you connect; agents never see repos you haven't shared.
• Least privilege. We request only the permissions a given task requires, and prefer fine-grained, short-lived credentials where the platform supports them.
• Revocable any time.You can disconnect a repository or revoke the integration from your code host's settings, and access ends immediately.

We store the minimum needed to run your company: account details, connected project content, session transcripts, and operational telemetry. Access to production data is restricted to a small set of authorized personnel on a need-to-know basis, and such access is logged.

• Least-privilege access. Internal access follows role-based controls and is reviewed regularly.
• No model training on your content.We don't use the contents of your private repositories or transcripts to train foundation models.
• Defined retention. Data is kept only as long as needed and then deleted or anonymized — see the Privacy Policy for specifics.

All traffic between you, the platform, and the services we depend on is encrypted in transit using modern TLS. Credentials and tokens for your connected integrations are stored encrypted and handled as secrets, never exposed in logs or transcripts.

We follow current best practices for key management and rotate secrets on a regular schedule. As the platform matures we continue to harden encryption at rest across our storage layers.

We welcome reports from security researchers and act on them quickly. If you believe you've found a vulnerability, please report it to us privately so we can investigate and fix it before it's disclosed publicly.

• Email security@stackbooster.io with the details and steps to reproduce.
• Give us a reasonable window to investigate and remediate before any public disclosure.
• Don't access or modify data that isn't yours, and avoid actions that could degrade the service for others.

We'll acknowledge your report, keep you updated on our progress, and credit researchers who help us keep Aria safe — with their permission.

For vulnerability reports, email security@stackbooster.io. For general security questions or to request our current sub-processor list and practices, use the contact page.