Skip to content

FAQ

No. Guardyx governs the tool-call path, not your agent logic. Point an MCP client at a Guardyx endpoint (Quickstart) or wrap calls with the SDK — either way your existing agent stays as-is.

No. Guardyx governs the runtime path agents use to act — the tool invocation path — not company compliance posture. There are no questionnaires, evidence collection, or trust-center automation.

Any agent that either runs in an MCP client or whose tool calls you can route through the SDK — LangGraph, LangChain, custom Python, AutoGen, CrewAI, OpenAI Agents SDK, Anthropic Claude Agent SDK, Google ADK, raw OpenAI tool calling, and more.

When a policy flags an action (for example, an external write above an auto-approve threshold), execution is blocked and routed to a human reviewer. The action executes only after approval, and the full decision is captured in the run trace.

Guardyx is in active development and onboarding design partners. Request early access from the home page.