Put Hud to work

Once the SDK runs in your services, Hud has runtime context for every function: it counts every call and learns each function's baseline. This page shows where to use it: before you change code, on your PR, when Hud flags an issue, and any time you have a question about production.

Before you change code

Check how the code runs in production before you change it. Ask Hud MCP before your coding agent writes the change, for example:

• "How does OrderService.create run in production, and what calls it?"
• "I'm changing OrderService.create. Which endpoints run it, and how much traffic do they carry?"

On your PR

Where the PR risk check is on for a repository, Hud reviews each pull request against how the functions it changes run in production. It posts a verdict, Safe to merge, Review required or High risk, with comments on the lines involved and a matching label. Read its findings before you merge, as you would a reviewer's. See PR risk check.

When Hud flags an issue

Hud opens an issue when a flow gets slower or starts failing, whether right after a rollout or over days.

• Where you hear about it: a Slack alert in your team's channel, or the Issues page in the web app. The alert names the flow and the function where the problem started. Select View issue to open it.
• What it tells you: the error rate or duration over time, the function where the problem started, and the forensics Hud captured, such as the request, the exception and outbound calls.
• What to do next: where Auto Remediation is on, select Run Remediation on the issue or its Slack alert, and Hud's agent finds the root cause and opens a PR with a proposed fix. Otherwise, select Prompt to Fix to hand the issue to your coding agent. The agent reads the evidence over Hud MCP and explains the root cause. Then ask it for the fix, and review the change like any other PR.

Any time

• In the web app: the Overview shows what's slowest, fails most or is called most, and what's trending up. The Explore pages list every service, endpoint, queue, custom flow and function, and a function's page shows its call graph. See Tour the web app.
• In your coding agent: ask about production whenever a question comes up, for example:
  • "Which functions make POST /orders slow in production?"
  • "Why are messages on the order-events queue failing?"
  • "Which functions in the payments service are worth optimizing?"
  • "Did /checkout get slower after the last rollout?"
• In your editor: Hudders above each function show its calls per day, its duration and its error rate. Click one to see which functions call it and which ones it calls.