Agent Configuration

Configure and customize your AI agents to match your specific needs. This section covers everything from writing effective prompts to enabling advanced tools.

Overview

Agent configuration determines how your AI agent behaves, what it can do, and how it interacts with users. The key components are:

  • System Prompts - Define the agent's role, personality, and behavior
  • Tools & Capabilities - Enable specific functionalities like code execution, web browsing, or file operations
  • Model Selection - Choose the right AI model for your use case
  • Permissions & Visibility - Control who can access and use your agents

Configuration Topics

Prompts and Instructions

Learn how to write effective system prompts that guide your agent's behavior.

You'll learn:

  • How to structure system prompts
  • Prompt engineering best practices
  • Common prompt patterns
  • Examples for different use cases

→ Read Prompts and Instructions Guide

Tools

Enable and configure tools that extend your agent's capabilities.

Available tools:

  • Bash - Execute shell commands
  • Filesystem - Read and write files
  • Web Search - Search the internet
  • Code Execution - Run code in isolated environments
  • MCP Servers - Connect to external services

→ Explore Tools Documentation

Quick Start

Basic Agent Configuration

{
  "name": "Customer Support Agent",
  "role": "Support Specialist",
  "systemPrompt": "You are a friendly customer support agent...",
  "model": "claude-3-5-sonnet-20241022",
  "advanced_tools": ["read_webpage", "exa"],
  "visibility": "organization"
}

Advanced Configuration

For production agents, consider:

  1. Detailed System Prompts - Include examples, constraints, and edge cases
  2. Tool Selection - Only enable tools the agent actually needs
  3. Safety Guidelines - Add guardrails for sensitive operations
  4. Testing - Thoroughly test configurations before deployment

Best Practices

Writing System Prompts

✅ Do:

  • Be specific about the agent's role and responsibilities
  • Include examples of desired behavior
  • Define clear boundaries and limitations
  • Use consistent tone and style

❌ Don't:

  • Make prompts too vague or generic
  • Assume the agent knows your business context
  • Skip testing with real user queries
  • Forget to update prompts as requirements change

Tool Configuration

✅ Do:

  • Enable only necessary tools (principle of least privilege)
  • Test tools individually before combining
  • Document why each tool is needed
  • Monitor tool usage in production

❌ Don't:

  • Enable all tools "just in case"
  • Skip security review for bash/filesystem tools
  • Forget to set appropriate permissions
  • Ignore tool usage patterns

Common Patterns

Customer Support Agent

{
  systemPrompt: `You are a helpful customer support agent for [Company].

Your responsibilities:
- Answer product questions accurately
- Troubleshoot common issues
- Escalate complex problems to human support
- Maintain a friendly, professional tone

Knowledge base:
- Product documentation in /docs
- FAQ in /knowledge/faq.md

Constraints:
- Never share internal information
- Don't make promises about features
- Always confirm before taking actions`,

  advanced_tools: ["read_webpage", "filesystem"]
}

Code Review Agent

{
  systemPrompt: `You are a senior developer reviewing code for quality and best practices.

Focus areas:
- Code quality and readability
- Security vulnerabilities
- Performance issues
- Test coverage
- Documentation

Review checklist:
1. Check for obvious bugs
2. Verify error handling
3. Look for security issues
4. Assess code organization
5. Suggest improvements`,

  advanced_tools: ["bash", "filesystem"]
}

Research Agent

{
  systemPrompt: `You are a research assistant that finds and synthesizes information.

Research process:
1. Understand the research question
2. Search for relevant sources
3. Verify information accuracy
4. Synthesize findings
5. Cite sources

Guidelines:
- Prioritize authoritative sources
- Cross-reference information
- Note any contradictions
- Provide source links`,

  advanced_tools: ["exa", "read_webpage"]
}

Next Steps

Ready to configure your agent? Start with these guides:

  1. Prompts and Instructions - Write effective system prompts
  2. Tools - Enable advanced capabilities
  3. Troubleshooting - Fix common configuration issues

Need Help?