Skip to main content
Automagically applies layered controls across identity, change governance, execution, AI safety and auditability. Every runbook is authenticated, permissioned, approved, tested, tied to a ticket and logged before and during execution. Use this page to answer a security review or a client questionnaire about Automagically.

​
Controls at a glance

​
Access and identity

Authentication. Automagically runs inside the Thread platform and uses the same sign-in. Users must sign in to Thread to build or run any runbook. There is no anonymous or standalone access path. For Thread’s certifications and audit reports, see the Thread Trust Center. Role-based access control (RBAC). Granular permissions govern who can build, approve and run runbooks, and which clients each user can reach. Build, approval and run rights are assigned independently, so you can separate the person who writes a runbook from the person who approves it. To limit Automagically to some of your clients, see Automagically: Client Access.

​
Execution and change governance

Ticket-bound execution. Every runbook run must be linked to a ticket. Nothing runs without a documented reason, and every run traces back to a client request or incident. Approval gates. A new runbook must pass code review and a successful staging test before it can run. Any change to an existing runbook, or to one of its steps, needs approval again before it goes live, so unapproved code never reaches a client environment. For how approvals work in practice, see Approving runbooks.

​
Code quality and AI safety

Best-practice validation. Each runbook is checked against coding best practices while it is being built. Automated Pester tests run before a build is released for approval, so reviewers only see code that has already passed testing. LLM guardrails. Every LLM call goes through Thread’s AI gateway (Portkey) and is governed by its filtering and control policies. The build process adds further guardrails against misuse and to enforce compliance requirements. Portkey is listed on Thread’s list of sub-processors. No training on customer data. No customer or client data that Automagically processes is used to train LLMs.

​
Audit and logging

Full audit trail. The build agent’s chat history is kept for each runbook, recording how and why it was created. Every approval and every run is recorded in the audit history: who did what, when, and against which ticket and client.