FAQ
Short answers to common questions.
In shadow mode, calls don't wait for Leap360 except for a kill switch check, at most once a second. In enforce mode, allowed calls are decided in your process in about 0.7 ms. Calls a policy would block or hold go to Leap360, which stays authoritative.
Your agents keep running. Only actions at or above the failClosedAt risk (default critical) are blocked until Leap360 is reachable again.
New projects start with a 7-day window. The project stays in shadow mode after that until a person switches it to enforce. You can start a new window of 1 to 30 days.
No. Those actions need a person signed in to the console. An API key can only make governance stricter.
Tool calls from OpenAI, Anthropic, Gemini and Bedrock models are read automatically. The code scan recognises the Vercel AI SDK, OpenAI Agents, Mastra, LangChain, MCP servers and Python tool decorators. Any MCP server can be governed with the MCP proxy.