Built for your production services
Gazebo holds the API keys to your production infrastructure. We take that seriously. Here's exactly how we protect your credentials, your agents, and your services.
Scoped agent identity and credential release
Each agent gets a Gazebo token and permission to retrieve selected service credentials. Method checks use the method declared by the agent at retrieval time; they do not enforce subsequent provider calls. Use narrowly scoped provider keys or IAM permissions to restrict actual service actions.
Encrypted credential vault
API keys and secrets are encrypted with AES-256-GCM before being written to the database, with the encryption key stored separately. Dashboard connection views do not show stored values. Authorized agent API and MCP requests return the actual provider credential to the runtime; a third-party host running that agent receives it too.
Approval controls for agent access
Agent Consent Flow records a user's approval of requested scopes through a separate authorization protocol. Its tokens are not automatically interchangeable with runtime agent tokens. SDK/CLI setup is currently manual; do not assume a consent grant gates every credential retrieval or subsequent provider action.
Credential-access audit log
Credential requests from authenticated agents record the agent, service, declared method, outcome and time. These events describe retrieval, not later provider actions. Token authentication failures do not create credential-access records; use provider logs to investigate downstream activity.
Authentication & session security
Passwords are hashed using PBKDF2-SHA512 with 100,000 iterations and a random salt — never stored in plaintext. Sessions use HTTP-only, server-side cookies that cannot be accessed by JavaScript. Magic link tokens are single-use and expire after 1 hour.
API and MCP access
Programmatic access via the REST API or MCP endpoint requires a Bearer token. Full-account tokens grant broad access; agent tokens are scoped to specific permitted services only. Both can be revoked at any time from your account settings. Every request is authenticated before any data is accessed or any action is taken.
Request protection
Gazebo validates credential requests against each agent's service and declared-method policy. It returns the configured provider credential, not a newly narrowed provider key. Provider permissions remain the enforcement boundary after retrieval; Gazebo does not proxy subsequent service calls.
One-click revocation
Revoking an agent blocks future credential retrieval with its Gazebo token without changing the provider key. It does not invalidate a key already retrieved or undo provider actions. If a key may have leaked, revoke or rotate it at the provider; this can affect other systems using that key.
Infrastructure
Gazebo is hosted on Replit's cloud infrastructure, built on Google Cloud Platform. All data is stored in a managed PostgreSQL database. Traffic between your browser and Gazebo is encrypted in transit via TLS 1.3.
Database backups run automatically. The credential vault encryption key is stored separately from the credential data and is never rotated without a migration plan.
Rate limiting is applied to agent credential requests to prevent abuse.
Vulnerability disclosure
If you discover a security vulnerability in Gazebo, please report it responsibly. Email us at [email protected] with a description of the issue and steps to reproduce it. We aim to acknowledge reports within 48 hours and resolve confirmed vulnerabilities within 14 days.
Please do not publicly disclose vulnerabilities before we've had a chance to investigate and fix them. We appreciate responsible disclosure and will credit researchers who help us improve security.
Worried a config file already has raw credentials in it?
Free scanner