GateBolt is a preflight approval and provenance layer for AI coding agents. Before an agent touches code, it submits its plan to GateBolt, where a deterministic policy engine evaluates the planned files and task against company policy and returns approved, requires_review, or blocked — along with a risk level, reasons, required approvers, and an audit trail.
It doesn’t write code, run the agent, or scan PRs after the fact. Its value is the layer before code changes: agent identity, planned scope, policy decisioning, human approval, and provenance.