Configure Behavior
Dieser Inhalt ist noch nicht in deiner Sprache verfügbar.
Behavior configuration defines how your agent responds to users. Through the system prompt, you control the agent’s personality, tone, knowledge boundaries, and what it should do when it doesn’t know the answer.
The system prompt
Section titled “The system prompt”The system prompt is a set of instructions the agent follows in every conversation. It’s the most powerful tool you have for shaping behavior.
Where to find it
Section titled “Where to find it”- Select your agent from the Agents list.
- Go to Agent Settings or Configure Behavior in the sidebar.
- Edit the System Prompt field.
What to include
Section titled “What to include”A good system prompt covers these areas:
| Area | Example |
|---|---|
| Identity | ”You are a customer support assistant for Acme Corp.” |
| Tone | ”Be friendly, concise, and professional.” |
| Knowledge boundaries | ”Only answer questions using the provided knowledge base. Do not make up information.” |
| Fallback behavior | ”If you don’t know the answer, say: ‘I’m not sure about that. Would you like me to connect you with our team?’” |
| Handoff rules | ”If the user asks to speak with a human, transfer the conversation to an operator immediately.” |
| Restrictions | ”Never discuss competitor products. Do not share internal pricing or discount codes.” |
| Language | ”Always respond in the same language the user writes in.” |
Example system prompt
Section titled “Example system prompt”You are a helpful support assistant for Acme Corp, an online storethat sells outdoor gear.
Rules:- Answer questions using only the knowledge base provided.- Be friendly, concise, and professional.- If you don't know the answer, say so honestly and offer to connect the user with a team member.- Never invent product details, prices, or policies.- If the user asks to speak with a human, transfer immediately.- Respond in the same language the user writes in.Tips for effective prompts
Section titled “Tips for effective prompts”- Be specific — vague instructions like “be helpful” don’t shape behavior. Tell the agent exactly what to do in common scenarios.
- Set boundaries first — start with what the agent should NOT do, then describe what it should do.
- Test iteratively — write a prompt, test it in the Playground, refine, and repeat.
- Cover edge cases — what should the agent say when asked about pricing? Competitors? Off-topic questions?
- Keep it concise — long prompts can dilute important instructions. Aim for clarity over completeness.
Behavior beyond the prompt
Section titled “Behavior beyond the prompt”The system prompt works together with other settings:
- Model selection — different models have different strengths. See Model Selection.
- Temperature — controls how creative or deterministic the responses are. See Agent Settings.
- Knowledge base — the quality and coverage of your documents directly affects response quality. See Knowledge Base.
- Triggers — automate specific behaviors based on conversation conditions. See Triggers.
Common patterns
Section titled “Common patterns”Strict knowledge-only agent
Section titled “Strict knowledge-only agent”Only answer questions based on the knowledge base. If the information is not in the knowledge base, say “I don’t have information about that” and offer to connect with support.
Conversational sales agent
Section titled “Conversational sales agent”You are a sales assistant. Recommend products from the catalog based on the user’s needs. Ask follow-up questions to understand what they’re looking for. If they’re ready to buy, share the product link.
Multilingual support agent
Section titled “Multilingual support agent”Detect the user’s language and respond in the same language. If you’re unsure, default to English. All knowledge base content is in English — translate your answers naturally.
Handoff-focused agent
Section titled “Handoff-focused agent”Answer simple questions from the knowledge base. For anything complex — billing, refunds, complaints, or technical issues — immediately offer to connect the user with a team member.
Next steps
Section titled “Next steps”- Model Selection — choose the right model for your use case.
- Agent Settings — configure temperature, token limits, and more.
- Playground — test your prompt in real time.