An autonomous safety auditor for AI coding agents that detects scope violations, exposed secrets, destructive commands, and deployment risks before execution.
Scope Guard is an agent-safety control plane and callable software service that evaluates code changes, shell commands and deployment plans before an autonomous coding agent acts.
Its deterministic policy engine detects cross-project scope violations, destructive commands, protected-resource access, unsafe infrastructure changes and approval-sensitive operations. It returns structured ALLOW, BLOCK or REQUIRE_APPROVAL decisions, together with risk explanations, audit evidence, validation requirements and rollback guidance.
Scope Guard enables development teams to benefit from fast autonomous agents without surrendering control over production systems.
Scope Guard entered the OKX.AI Genesis Hackathon with a functional agent-safety control plane originally developed as a developer-tools project. The existing foundation includes deterministic policy enforcement, project-boundary protection, approval gates, task-scoped rollback, hash-chained audit reporting and a 32-scenario safety evaluation suite.
During the OKX.AI Genesis Hackathon, the project is being adapted into a publicly callable Agent Service Provider for the Software Services category. The hackathon work focuses on exposing the policy engine through a read-only A2MCP-compatible API, producing structured ALLOW, BLOCK and REQUIRE_APPROVAL responses, preparing marketplace-ready service documentation, validating the hosted Railway endpoint and creating an OKX-specific demonstration workflow.
The adaptation preserves the existing safety architecture while converting it from a standalone developer control plane into a reusable service that other AI agents can call before executing sensitive code or infrastructure operations.
Scope Guard is currently bootstrapped and has not raised external capital. Product development, testing and infrastructure have been funded directly by the founder.
The immediate focus is validating demand through the OKX.AI marketplace, collecting feedback from developers using autonomous coding agents and securing pilot integrations with software teams operating shared or production-sensitive infrastructure.
The planned commercial model combines a free entry-level safety analysis service with paid advanced audits, organisation policy packs, signed compliance reports and enterprise deployment controls. Strategic investment and ecosystem partnerships will be considered after demonstrating repeat usage, measurable risk reduction and a clear path to recurring API and enterprise revenue.