Agents
Write clear system instructions
Write clear system instructions
Give your agents a focused, specific purpose. Avoid vague instructions — the more precise you are about what the agent should and should not do, the more consistent its behavior will be.
Scope each agent to a single domain
Scope each agent to a single domain
Create separate agents for separate use cases rather than building one agent that does everything. Focused agents are easier to maintain and perform better.
Test before deploying
Test before deploying
Use the conversation preview to test your agent thoroughly before making it live. Try edge cases and unexpected inputs to see how it handles them.
Knowledge Base
Keep content up to date
Keep content up to date
Regularly review and update your knowledge base documents. Outdated information leads to incorrect agent responses.
Use clear, well-structured documents
Use clear, well-structured documents
Agents retrieve content more reliably from documents with clear headings, concise sections, and consistent formatting.
Only include relevant content
Only include relevant content
Avoid adding documents that are not relevant to your agent’s purpose. Irrelevant content can confuse retrieval and degrade response quality.
Security
Apply least-privilege roles
Apply least-privilege roles
Assign team members only the permissions they need. Use Roles & Permissions to create scoped roles rather than giving everyone admin access.
Rotate publishable keys periodically
Rotate publishable keys periodically
Regularly rotate your API keys and revoke any that are no longer in use to reduce exposure.