Handbook answering is retrieval plus restraint, not deep reasoning, so it does not need your most expensive model on every mention. OpenCode is model-agnostic, which means starting modest and moving up only if answers come back vague or citations start going missing. The Notion wiring and the refusal rules stay untouched while you experiment.
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.
How much parental leave do I get?
Priya Raghavan
@hr our baby is due on 14 March, how much parental leave do I get and is any of it paid? Also is there a date I have to tell my manager by?
Using this template drops you into a guided setup. It asks exactly this, nothing else:
Connect Slack
One sign-in. The agent acts through your account, scoped to what this template uses.
Connect Notion
One sign-in. The agent acts through your account, scoped to what this template uses.
Company name
Your company name as your team says it in Slack, so the assistant sounds like it works there.
Policy notes
Where your policies live in Notion, page or database by page, covering leave, benefits, expenses, and working hours, plus any rule your team follows that nobody has written down yet. The assistant only ever answers from what you list here.
Hr contact
Who takes what a document cannot answer, written exactly as you want it said in Slack: a person's handle, an HR channel, or an email. If different people cover different things, say which is which.
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.
Two sentences and a page link is a light generation task. Choosing the model yourself is what keeps an always-on assistant affordable.
Your policy notes, HR contact and refusal list live in the agent's variables, so changing what runs underneath does not disturb any of them.
That is usually one of two fixes: sharpen the policy notes so the assistant knows which pages are authoritative, or move the sub-model up a tier. Both are edits on the same agent, and you can compare before and after on the staged parental leave question rather than on a real employee.
Free to start. Guided setup, a test run against staged conversations, and it's live.