tool-approval

Runtime guardrails for AI agent MCP tools. Blocks dangerous tool calls at runtime with a default-deny policy proxy.

Minimal shell over Claude: Blazor UI + Claude Agent SDK sidecar with browser-based tool approval (Allow/Deny on every tool call). No injected prompt, no governance - fork it to build your own workflow.

The form renderer for MCP tools, A2A agents, and human-in-the-loop AI. One component. Any JSON Schema. Beautiful forms. Under 1ms.

[READ ONLY] Tool approval guard middleware for Laravel AI agents.

Runtime risk guard for OpenClaw: hook-based tool interception, approval workflows, prompt injection defense, and auditable local security controls.

Three-tier intelligent permission gate for AI agents — automated approval with hash-based trust, AI evaluation, and full audit ledger

Policy-aware AI permission reviewer for OpenCode. Evaluates tool and command requests, then allows once, denies with feedback, or escalates to the user.