Security
Security at every layer of agent execution
Prodedify is built so AI agents can write and run code without putting your repository, your secrets, or your data at risk.
We took security as a non-negotiable from day one. Agents execute in isolated cloud sandboxes, data is encrypted in transit and at rest, every action is recorded in an exportable audit log, and your content is never used to train models. The result: AI that actually ships code, with controls your security team can defend.
Sandboxed agent execution
Every agent runs inside an isolated cloud sandbox built on lightweight microVMs — strong process and network isolation between the agent and your environment.
- Per-run isolated sandboxes (Firecracker microVMs via E2B)
- Restricted network access by default; explicit allow-listing for outbound calls
- Resource limits prevent runaway processes from affecting other work
Encryption in transit and at rest
Your data is encrypted on the wire and on disk, using modern, widely-reviewed cryptography — and we keep the surface area small on purpose.
- TLS 1.2+ for all traffic between client, API, and agents
- AES-256 encryption at rest for stored data
- Secrets stored in a managed secrets layer, not in plaintext
Authentication & access control
OAuth and SSO for people, scoped API keys for machines, and role-based permissions so the right humans (and the right agents) can act only where they should.
- OAuth and SSO (SAML / OIDC) for human sign-in
- Scoped API keys with per-key permissions and rotation
- Role-based access control down to workspace, project, and resource
Audit logging
A complete, queryable record of every action taken — by people, by APIs, and by autonomous agents — so reviews are evidence-based and incidents are traceable.
- Every user, API, and agent action recorded with actor and context
- Exportable audit logs for your SIEM / data warehouse
- Retention configurable to match your compliance program
How we handle your data
Your code, specs, and prompts are yours. We do not use customer content to train models, and we keep a clear, written boundary between operating the service and improving it.
- Customer content is not used to train AI models
- Workspace data is logically isolated and tenant-scoped
- Clear, documented data-handling boundaries for support and operations
Responsible disclosure
Security issues happen. When you find one in Prodedify, we want to hear about it — quickly, privately, and with a clear path to a fix.
- Private channel for security researchers to report issues
- Acknowledged, triaged, and tracked to resolution
- Coordinated disclosure once a fix is in customers’ hands
Security, answered
Each agent run executes inside its own cloud sandbox — a lightweight microVM with restricted network access and resource limits. Nothing reaches your repository without a pull request you approve, so a human always has the final say.
No. Customer content is not used to train AI models. Your code, specs, and prompts remain yours, with a clear written boundary between operating the service and improving it.
TLS 1.2 or higher for all traffic between client, API, and agents. AES-256 for data at rest. Secrets are kept in a managed secrets layer, never in plaintext.
Each API key is bound to a workspace and a permission scope. You can issue keys for specific automations, rotate them on a schedule, and revoke them instantly — without disrupting the rest of your integration.
Yes. Audit logs are exposed via API so you can stream them into your SIEM or data warehouse for long-term retention, alerting, and analysis alongside the rest of your security telemetry.
Email [email protected] (or contact us via the form). Reports are triaged quickly and tracked to a fix, with coordinated disclosure once customers are protected.
Explore the platform
Bring your security review
Run Prodedify with your existing identity provider, audit pipeline, and policies. We will work through your review with you.