Welcome

Meet Hud

Coding agents write a change in minutes. Getting it to production still runs on people. They review the PR, watch the rollout and dig through logs when something breaks, because the agent has never seen the code run. Hud gives you and your coding agents runtime context for every change, so each of those steps runs on evidence.

Hud's runtime code sensor runs inside each backend service and learns its baseline. For each function, Hud knows:

• How often it runs and how long it takes, for each endpoint, queue and caller.
• What calls it and what it calls, including databases and other services.
• Which exceptions it throws, even the ones your code catches.
• How all of that changed with the latest deployment.

When something goes wrong, Hud has already captured the request, the call graph and the exception with its stack trace. You don't pick metrics or write thresholds.

The same runtime context goes wherever your code changes:

• Before the merge, in a risk check that rates each PR, so safe ones can merge without waiting for a person.
• As the new version takes traffic, in a check against earlier versions.
• When a function regresses, as the evidence your coding agent needs to fix it.

Hud works with whichever coding agent your team uses, like Claude Code, Cursor or Codex.