A tour of the app, page by page.
Overview
Where you land. Requests over the last 24 hours, the allow/deny split, average policy evaluation time, and how many latches are live. Below that, a live activity feed of the requests happening right now, filterable to just the denials or just the slow ones.
If the deny rate spikes, this is where you notice.
Secrets
The real credentials. Bind one, and Latch encrypts it and never gives it back. Each secret shows the latches derived from it. See Secrets.
Latches
The scoped tokens you hand out. Grouped into My latches, Examples, All, Enabled, and Delegated (children extended from a parent).
From a latch you can simulate it, clone it, extend it, disable it, or turn on request and response body logging. See Latches.
Simulate
Dry-run a request through a latch's real pipeline without calling the upstream. Pre-built scenarios, or a custom request of your own. See Simulate.
Activity
The audit log. Every request, allowed or denied, with the filter that decided and why. Exports to CSV. See Activity.
Connect
Generates the config an agent needs for a given latch: MCP setup for Claude Code, Claude Desktop, Cursor and OpenClaw, environment variables, and an agent prompt describing what that latch can do. See Connect an agent.
Everywhere
- Command palette (
⌘K) to jump to any latch, secret, or page. - Theme: system, light, or dark.