How Knowledge Impacts Agent Responses

How GRUP Agents Use Project Knowledge

• Every message your agent sends is based on what it knows

• It scans your project content to find relevant answers to questions

• It uses tone and personality settings to match how things are said, but not what is known

Without strong knowledge, your agent becomes generic or misinformed. With quality input, it becomes an intelligent community assistant.

Good vs. Poor Responses (Examples)

Example 1: Token Utility Question

User: “What does this token actually do?”

• With poor knowledge: “This token supports the project.”

• With good knowledge: “$FLUX powers our DeFi lending protocol, giving users access to low-interest crypto loans and governance voting.”

Example 2: App Navigation

User: “Where can I use the mini app?”

• With poor knowledge: “Check our site.”

• With good knowledge: “You can access the Mini App directly at t.me/YourAppBot. Tap the 'Play Now' button to get started.”

What Happens When Knowledge Is Missing

If your knowledge base lacks details:

• The agent might generate answers that are too general

• It could skip the question or deflect with uncertainty

• It might link to unrelated or outdated resources

How to Improve Agent Answers

• Regularly update your knowledge base with new features, FAQs, and events

• Add clear explanations of your token, roadmap, and app flow

• Write like you're training a smart team member—not a search engine

• Use structured formatting (headers, bullets, bold keywords)

Testing Your Agent’s Understanding

Use the Live Chat tab to ask real user-style questions:

• “Where can I buy this token?”

• “What’s the use of staking?”

• “Is this project audited?”

If the responses are weak, improve the project knowledge first.