Agent Tools

Configure existing agents as tools for other agents

Overview

Agent Tools allows you to use already created agents as tools for other agents, creating an architecture of specialized agents that can be called as needed. This functionality offers:

• Reuse of specialized agents in different contexts
• Modularization of specific functionalities
• Intelligent orchestration between multiple agents
• Specialization of tasks by knowledge domain

Agents as Tools: Any agent created on the platform can be used as a tool by other agents. This allows creating hierarchies and functional specializations.

How to Configure Agent Tools

Step 1: Access Agent Settings

Navigate to settings

1. Go to the agents screen in the dashboard
2. Locate the agent that will be the “orchestrator” (that will use other agents as tools)
3. Click on the “Settings” icon (⚙️) on the agent card and then on “Edit”
4. You will be directed to the agent settings screen

Access Agent Tools section

1. In the settings screen, locate the “Agent Tools” section
2. Click on “Add” to add agents as tools
3. You will see the list of all available agents in your account

Only active and functional agents will appear in the selection list.

Step 2: Select Agents

View available agents

In the selection screen, you will see all agents created in your account:

Information displayed:

• Agent name and description
• Agent type (LLM, Task, Workflow, etc.)
• Status (active/inactive)
• Specialization or area of expertise

Select specific agents

How to select:

1. Check the checkboxes of the agents you want to use as tools
2. Consider the specialization of each agent
3. Avoid redundancies - don’t select agents with very similar functions
4. Think about hierarchy - the main agent should orchestrate the specialized ones

Selection criteria:

• ✅ Complementarity - agents with different specialties
• ✅ Reliability - agents already tested and functional
• ✅ Performance - agents with good response speed
• ❌ Avoid loops - don’t select the agent itself

Select agents that have complementary specialties to maximize efficiency.

Confirm selection

1. Review the selected agents
2. Confirm they make sense for the main agent’s context
3. Click on “Save” or “Confirm”
4. The selected agents will be available as tools

Usage Examples

Customer Service Agent with Specialists

Scenario: Customer Service Center

Main Agent: General Customer Service Representative

Agents as Tools:

• 🛒 Sales Agent - Specialist in products and prices
• 🔧 Technical Support Agent - Solves technical problems
• 💰 Financial Agent - Payment and billing issues
• 📦 Logistics Agent - Tracking and delivery

Workflow:

Customer: "My order didn't arrive and I want to cancel"

General Customer Service Representative:
1. Calls Logistics Agent → checks order status
2. Calls Financial Agent → processes cancellation
3. Responds to customer with complete solution

Research Agent with Specialists

Scenario: Academic Research

Main Agent: General Researcher

Agents as Tools:

• 📊 Data Analysis Agent - Processes statistics
• 📚 Literature Agent - Searches academic references
• 🧪 Methodology Agent - Suggests research methods
• ✍️ Writing Agent - Helps with scientific writing

Usage example:

User: "I need to research AI in education"

General Researcher:
1. Calls Literature Agent → searches relevant papers
2. Calls Analysis Agent → analyzes trends
3. Calls Methodology Agent → suggests approach
4. Compiles complete report

Development Agent with Specialists

Scenario: Software Development

Main Agent: Software Architect

Agents as Tools:

• 🎨 Frontend Agent - UI/UX specialist
• ⚙️ Backend Agent - APIs and databases
• 🧪 Testing Agent - Testing strategies
• 🚀 DevOps Agent - Deploy and infrastructure

Collaborative workflow:

User: "I want to create an e-commerce"

Architect:
1. Calls Frontend Agent → defines interface
2. Calls Backend Agent → plans API
3. Calls DevOps Agent → defines infrastructure
4. Calls Testing Agent → QA strategy
5. Delivers complete architecture

Best Practices

Agent Organization

Hierarchical structure

Organization principles:

Orchestrator Agent (Main):

• 🎯 Overall vision of the problem
• 🧠 Routing intelligence for specialists
• 📋 Consolidation of responses
• 🎭 Interface with the user

Specialist Agents (Tools):

• 🔬 Specific focus on one area
• ⚡ Quick and accurate responses
• 📚 Deep knowledge of the domain
• 🎯 Well-defined tasks

Maintain a clear hierarchy: one orchestrator agent and several specialists.

Configuration tips

For the Main Agent:

Instructions: "You are an intelligent orchestrator. 
Analyze the user's request and determine which 
specialist agent should be called. You have access to:
- Sales Agent (products and prices)
- Technical Agent (support and problems)
- Financial Agent (payments)

Always consolidate the specialists' responses 
into a single, coherent response for the user."

For Specialist Agents:

Instructions: "You are a specialist in [AREA]. 
Answer only questions related to your domain. 
Be precise, direct and technical. If the question is not 
in your area, inform that it's not your specialty."

Performance and Efficiency

Performance optimization

Optimization strategies:

• ⚡ Limit the number of agent-tools (maximum 5-7)
• 🎯 Specialize well each agent to avoid overlap
• 📊 Monitor latency of calls between agents
• 💰 Consider costs of multiple API calls
• 🔄 Avoid infinite loops between agents

Too many agent-tools can significantly increase response time and costs.

Monitoring

Important metrics:

• 📈 Usage rate of each agent-tool
• ⏱️ Average response time
• 💸 Cost per complete interaction
• ✅ Success rate of calls
• 🔄 Usage frequency of each specialist

Use analytics to identify underutilized agents and optimize configuration.

Troubleshooting

Common problems

Agent-tool not responding:

• Check if the specialist agent is active
• Confirm you have available API credits
• Test the agent individually first
• Check error logs in the dashboard

Inconsistent responses:

• Review the orchestrator agent’s instructions
• Improve the specialization of agent-tools
• Avoid overlapping responsibilities
• Test specific scenarios in isolation

Slow performance:

• Reduce the number of agent-tools
• Optimize instructions for more direct calls
• Consider using faster models
• Monitor communication bottlenecks

High costs:

• Analyze agent usage patterns
• Optimize instructions to reduce unnecessary calls
• Consider more economical models for simple tasks
• Implement cache for frequent responses

---

🤖 Agent Tools configured! Now you can create complex architectures of specialized agents working together to solve problems more efficiently!