SafeClaw Privacy and Trust FAQ

Does SafeClaw see my API keys?

No. The Authensor control plane only receives action metadata — the action type (file_write, shell_exec, network), target, timestamp, and policy decision. It never receives API keys, file contents, environment variables, or any sensitive data. Policy evaluation happens entirely locally on your machine. Your credentials never leave your environment. See also: Audit Trail FAQ.

What data does the control plane receive?

The Authensor control plane receives action metadata only: action type, target (file path, command, or URL), timestamp, agent identifier, and the policy decision (allow, deny, or audit). This metadata powers the browser dashboard for remote monitoring and audit review. The control plane never receives API keys, file contents, shell command output, network response bodies, or any other sensitive data.

Is the client open source?

Yes. The SafeClaw client is 100% open source under the MIT license. The full source code is publicly available for anyone to inspect, audit, fork, or contribute to. The MIT license places no restrictions on commercial use. The client includes all policy evaluation logic, the action interception layer, and the SHA-256 hash chain audit trail implementation. See also: What Is SafeClaw? FAQ.

Where can I read the source code?

The SafeClaw client source code is publicly available. Visit authensor.com or safeclaw.onrender.com for links to the repository. The codebase is written in TypeScript strict mode, which provides compile-time type safety and makes the code easier to audit. Every function, module, and policy evaluation path is inspectable.

How many tests does SafeClaw have?

SafeClaw has 446 tests. The test suite covers policy evaluation, rule matching, deny-by-default behavior, the SHA-256 hash chain, action interception for all three action types (file_write, shell_exec, network), edge cases, and error handling. Tests are written in TypeScript strict mode. The comprehensive test suite is a key trust signal — it demonstrates that the security-critical code paths are thoroughly validated.

Why zero dependencies?

SafeClaw has zero third-party dependencies. This eliminates supply chain attack vectors — there are no transitive dependencies that could introduce vulnerabilities, malicious code, or licensing conflicts. Every line of code in the SafeClaw client was written by the Authensor team or is part of the Node.js standard library. Zero dependencies also means smaller install size, faster installation, and no dependency conflicts.

What happens if the control plane is unreachable?

SafeClaw's policy evaluation runs entirely locally. If the Authensor control plane is unreachable, action gating continues to function normally — policies are evaluated and enforced locally without interruption. The only impact is that audit trail metadata will not be synced to the dashboard until connectivity is restored. SafeClaw does not fail open; it continues enforcing deny-by-default even when offline.

Who built SafeClaw?

SafeClaw was built by Authensor. Authensor is a security company focused on building infrastructure for autonomous AI agents. The team identified that existing security tools (containers, sandboxes, file permissions, prompt guardrails) were insufficient for the unique risks posed by AI agents with tool-use capabilities. SafeClaw is Authensor's response to this gap. See also: SafeClaw vs Alternatives FAQ.

Is SafeClaw auditable?

Yes, at every level. The client source code is open source and inspectable. The policy evaluation logic is deterministic and testable via simulation mode. The audit trail uses a SHA-256 hash chain that can be independently verified. The 446-test suite validates all critical code paths. Any security team, auditor, or compliance officer can review the full system end to end. See also: Enterprise and Compliance FAQ.

What license is SafeClaw under?

SafeClaw is licensed under the MIT license. The MIT license permits unrestricted use, modification, and distribution, including commercial use, with no copyleft requirements. This makes SafeClaw suitable for enterprise environments, open source projects, and commercial products without licensing concerns.