Prompt to Fix

Prompt to Fix hands an issue to your own coding agent. It gives you a prompt that names the issue and points the agent to the evidence Hud captured, so the agent starts from what happened in production instead of guessing from the code. Use it with the coding agent you already work with, such as Cursor, Claude Code or Copilot.

Before you start

Prompt to Fix works through Hud MCP: the prompt names the issue, and your coding agent pulls the runtime context it needs over Hud MCP. See MCP · Connect. Without Hud MCP, you can still use Prompt to Fix on error issues. Select standalone prompt in the Prompt to Fix drawer, and the evidence comes attached to the prompt.

Get the prompt

Open the issue in the web app, on its page or in its drawer on the Issues page.
Select Prompt to Fix.
Select the copy icon to copy the prompt. If you use Hud in Cursor, you can select Open in Cursor instead.
Paste it into your coding agent, in the repository of the service the issue is on.

The agent investigates and explains the root cause. Then ask it for the fix, and review the change like any other PR.

You can also get a prompt for a single forensic: in the forensics table on a service, endpoint, queue, custom flow or function page, select Get Prompt to Fix. Prompt to Fix isn't in Slack alerts or the IDE extension.

Prompt to Fix or Auto Remediation

Prompt to Fix works on any issue and keeps the investigation in your own agent. Auto Remediation runs an agent for you on eligible issues and opens a PR. When an issue isn't eligible for remediation, or you disagree with its result, Prompt to Fix is the manual path.