Leap360 docs · Beta
Govern your AI agents
Everything you need to connect your agents to Leap360, from the first governed tool call to production.
import { Leap360 } from 'leap360';
const leap = new Leap360({
apiKey: process.env.LEAP360_API_KEY,
projectId: 'my-app',
serviceName: 'deploy-bot',
});
// Checked against your policies before it runs, then recorded.
const deleteNamespace = leap.wrapTool(
'delete_namespace',
async ({ name }) => k8s.deleteNamespace(name),
);Connect
Choose how you connect
Use the SDK for agents you build, the MCP proxy for agents you don't, and the code scan to see what any AI in your code can reach.
SDK
Wrap the tools your own agents call. Each call is checked before it runs, decided locally in about 0.7 ms when it is allowed.
MCP proxy
Put Leap360 in front of an agent you don't control, such as Claude Code, Cursor or any MCP server.
Code scan
Find every LLM call and tool an AI can reach in your codebase, before the code ships.
Your path
From first call to production
Follow the steps in order, or jump to what you need.
- 1
Get started
- 2
Observe in shadow mode
- 3
Enforce
- 4
Operate
Help center
Get help
Troubleshooting
Fixes for the problems people hit most.
FAQ
Short answers about speed, uptime and pricing.
Security
What leaves your process, and how it is protected.
Changelog
What changed in each release.
Contact support
Email support@leap360.ai.
Open the console
See your agents, approvals and the kill switch.