Similar tickets

Find tickets that look like the one you're working on.

When you're working on a ticket and want to know "have we seen this before?", the Similar Tickets feature pulls up the closest matches from your archive.

How it works

Click 🔍 Find similar tickets in the ticket sidebar.

  1. We run a MySQL full-text search on the current ticket's subject + description, scoped to the same workspace, to build a candidate set of up to 20 tickets.
  2. We send Claude the current ticket + the 20 candidates and ask it to score each candidate's similarity from 0 (unrelated) to 1 (basically the same issue).
  3. Claude returns the ranked list. We show the top 5 with their scores above 0.4.

The result renders as a list of clickable tickets in the sidebar with their reference, subject, status, and similarity score. Click a row to open that ticket in a new tab.

What "similar" means

The score weighs:

  • Subject overlap — same topic words.
  • Description overlap — same problem described in similar terms.
  • Resolution pattern — already-resolved tickets that match get highlighted; you can see how the previous one was handled.

It's not just keyword matching — the LLM groups by semantic intent. "I can't log in" and "Password reset email never arrives" might score 0.7 even with no shared keywords.

When it's most useful

  • Recurring issues — you want to find the canonical resolution.
  • Outage detection — five tickets in an hour about the same thing means it's not just one customer.
  • Onboarding new agents — they can see how veterans handled the same question.

When it's worst

  • Brand-new workspaces with few resolved tickets to match against.
  • Highly customer-specific issues (account #42's specific data) where similarity doesn't transfer.

Cost

The FTS candidate set is free. The LLM ranking is roughly 3000 input + 300 output tokens — about $0.005 per click.

Off-switch

Settings → AI → uncheck Similar tickets. The button returns "feature disabled".