Local-first tool collection

Small tools for
safer AI agents.

Each utility runs in the browser, explains its limits, and targets a specific configuration or security decision.

Why separate tools?

MCP configuration, OAuth discovery, repository instructions, and Agent Skills have different trust boundaries. A single generic score would hide those differences. Each site keeps its checks narrow and distinguishes static evidence from claims that require live or human review.

Privacy model

Pasted configuration and instruction content is processed locally. The tools do not require an account and do not send scan inputs to a backend.