The Core Agent works great out of the box. But you can tune how it behaves so it fits your workspace — its tone, its rules, which tools it may use, and what documents it can read.
Describe, in plain words, how the agent should act — its tone and any rules. Example: “Always answer in a friendly, concise way. Use our company’s product names. Never share internal pricing.” Think of it as the agent’s standing orders.
Model — its brain
Choose which AI powers the agent. Faster models for everyday work, smarter models for harder tasks. See Choosing a model. The default is a solid choice.
Tools — what it's allowed to do
Turn on the abilities the agent may use — like search or a connected service. Many tools come from your integrations. Give it only what it needs.
Knowledge — what it can read
Add documents so the agent answers from your information — policies, product docs, notes. See Ask your own documents.
Give the agent only the tools it needs. More tools is not better — it can slow the agent down or confuse it. Add one, test, then add another if needed.
After you adjust a setting, open a new chat in the Agents tab and try a task. Tweak the instructions until the agent behaves exactly how you want. Small, clear rules work best.
Changes to the Core Agent apply across your workspace. If several people share the workspace, agree on the instructions together so everyone gets consistent results.
