Classify per the triage rules and nothing else, and if the case fits nothing, say so plainly and route it to the default. That instruction protects your queue from invented categories, and how well it holds is mostly a function of the model. OpenCode makes that a field you can tune per org.
Run a staged conversation — no account needed. The agent handles it for real while a simulated world answers its tool calls; nothing touches real accounts, and nothing is actually sent.
Yuki Tanaka
Invoices exported with the wrong tax rate since Friday
Emaily.tanaka@murakami-denki.co.jp
NoteCase opened from the web form. Account Murakami Denki KK, Business plan, contact Yuki Tanaka, Finance Systems Lead, priority set by the customer to Medium, product field Billing, version 5.2. Description: since the update on Friday every invoice we export as CSV shows 8 percent consumption tax instead of 10 percent, while the same invoice in the web view is correct. It affects all 1,200 invoices we exported this morning for the July close and our accounting team has stopped the run. Nothing changed on our side apart from installing the update.
Using this template drops you into a guided setup. It asks exactly this, nothing else:
Connect Salesforce
One sign-in. The agent acts through your account, scoped to what this template uses.
Connect Slack
One sign-in. The agent acts through your account, scoped to what this template uses.
Triage rules
How you sort cases - your categories and what each one means, which queue or owner takes it, and the exact fields the agent may set at triage with the values it may use. Anything not written here is left for a human.
Escalation criteria
The cases that need a person immediately and a Slack ping - the severity, account tier, product areas, or wording that qualifies, and who should see them.
Runs on OpenCode
Preselected for this page — connect your OpenCode account during setup, or switch to NoClick's built-in models with one click.
Watch it handle a test run
A staged conversation against a simulated world — then it’s live.
A model that would rather guess than admit a gap creates a queue nobody trusts. Test this behaviour whenever you swap models.
High-volume support teams can trade model strength against run count deliberately, since the workflow itself does not change.
Run Vague Case, where the whole description is that it says error when I try to log in. Correct output classifies conservatively, admits the cause is unknown, and lists the exact questions the agent should ask. A model that diagnoses an authentication bug from that sentence will do the same on real cases.
Free to start. Guided setup, a test run against staged conversations, and it's live.