Unlocking Your Agent's Power: Integrating APIs and Custom Tools with bots.do

An autonomous AI agent is more than just a powerful language model; it's a digital worker designed to accomplish goals. While its "brain" can reason, plan, and understand complex instructions, its true power is unleashed when it can act. On the bots.do platform, action is enabled through tools.

Tools are the hands and feet of your AI agent. They are the bridge between the agent's digital mind and your business's operational reality. They transform your agent from a brilliant consultant into a proactive team member that can execute tasks, manipulate data, and drive real-world outcomes.

This post will guide you through why tools are essential and how you can equip your bots.do agents with custom APIs to build a truly autonomous digital workforce.

From Thinker to Doer: Why Tools are Non-Negotiable

Imagine hiring a new employee. They're incredibly smart, can strategize all day, but have no computer, no phone, and no access to your company's software. They can't do much, right? A tool-less AI agent is in the same position.

The bots.do agentic workflow platform is built on a simple but powerful premise: an agent uses AI to understand a goal, plan actions, and use tools to execute those actions. The "use tools" part is what separates a bots.do agent from a simple chatbot.

• A chatbot might say: "To resolve your issue, please check our knowledge base."
• A bots.do agent will do: It will autonomously use the knowledgeBase.search tool, analyze the results, and use the email.send tool to deliver a tailored answer directly to the customer.

By giving an agent access to your APIs, you are granting it the ability to perform the same digital tasks as a human employee, turning repetitive business services into scalable software.

Connecting Your World: How bots.do Agents Use Tools

At its core, a "tool" for a bots.do agent is an API endpoint that it can call to perform a specific action. You grant these capabilities when you create your bot, as shown in our basic example:

import { bots } from '@do/sdk';

// Create a bot to handle customer support inquiries
const supportBot = await bots.create({
  name: 'CustomerSupportAgent',
  goal: 'Resolve customer support tickets by analyzing ticket content, searching the knowledge base, and providing a helpful response.',
  tools: ['knowledgeBase.search', 'tickets.update', 'email.send']
});

console.log('New Support Bot Deployed:', supportBot.id);

In this code, the tools array is the most critical part. It's the agent's toolbox. Here's what those tools might represent:

• knowledgeBase.search: An internal API endpoint that takes a query and returns relevant articles from your company's documentation.
• tickets.update: A wrapper for your Zendesk, Jira, or Salesforce Service Cloud API that allows the agent to update a ticket's status or add a comment.
• email.send: An integration with SendGrid, AWS SES, or a similar service to send programmatic emails.

This "Business-as-Code" approach allows you to define your operational capabilities and securely grant them to your digital workforce.

Practical Walkthrough: Building a Lead Qualification Agent

Let's build a more advanced agent from scratch. Our goal is to automate the tedious process of qualifying new sales leads.

Step 1: Define the Agent's Goal

First, we need a clear, actionable goal for our agent.

Goal: "To qualify new inbound leads by enriching their data with company information and updating the corresponding record in our CRM."

Step 2: Identify the Necessary Tools

Think about the steps a human sales development representative (SDR) would take:

1. Look up the lead's company to find its size, industry, etc.
2. Check the CRM to see if the lead's company already exists.
3. Create or update the contact record in the CRM with the enriched data.
4. Tag the lead as "Qualified" or "Needs Review" based on the data.

This translates into a need for the following tools:

• A data enrichment API (like Clearbit, Hunter, or an internal database).
• A CRM API for searching, creating, and updating records (like HubSpot or Salesforce).

Step 3: Expose Your Tools as Secure APIs

You don't give the agent raw API keys. The best practice is to create simple, secure wrapper functions or API endpoints that your bots.do agent can call. These endpoints handle authentication and abstract away complexity.

For example, you might create:

• POST /api/enrichment/company_info
• GET /api/crm/search_contact
• POST /api/crm/update_contact

Step 4: Register the Tools with Your New Bot

Now, we create our LeadQualificationAgent and give it access to the endpoints we just designed.

import { bots } from '@do/sdk';

const leadQualifierBot = await bots.create({
  name: 'LeadQualificationAgent',
  goal: 'Qualify new inbound leads by enriching company data from publicly available sources and updating the contact record in our CRM.',
  tools: [
    'enrichment.getCompanyInfo', 
    'crm.searchContactByEmail', 
    'crm.updateContactRecord'
  ]
});

console.log('New Lead Qualifier Bot Deployed:', leadQualifierBot.id);

Step 5: Put the Agent to Work

With the agent deployed, you can now trigger it via a simple API call whenever a new lead comes in. When the agent receives a task like "Qualify lead: contact@newcompany.com", its agentic workflow kicks in:

1. Plan: The agent reasons: "My goal is to qualify this lead. First, I should see if they are in the CRM. Next, I'll enrich their data. Finally, I will update the CRM record with all the information."
2. Execute:
   • It calls crm.searchContactByEmail({ email: 'contact@newcompany.com' }).
   • Based on the result, it calls enrichment.getCompanyInfo({ domain: 'newcompany.com' }).
   • It synthesizes the information and calls crm.updateContactRecord({ ... }) with all the new data and an updated qualification status.

This entire multi-step process happens autonomously in seconds, turning a 15-minute manual task into an on-demand, scalable service.

Best Practices for Creating Agent Tools

• One Tool, One Job: Create granular, atomic tools. Instead of a single manageCRM tool, build specific ones like crm.search, crm.update, and crm.create. This gives the agent more precise control and a better chance of success.
• Descriptive Naming is Key: Your tool names and descriptions are the primary way the agent understands what it can do. Be clear and descriptive. The agent's LLM uses this metadata to choose the right tool for the job.
• Prioritize Security: Always use wrapper functions and secure API gateways. Never expose raw credentials or give agents overly broad permissions. The bots.do platform provides a secure environment, but good tool design is your first line of defense.
• Start Simple and Iterate: Begin by giving your agent one or two essential tools. Observe its performance, refine the tools and their descriptions, and gradually expand its capabilities.

Build Your Digital Workforce Today

An AI agent's intelligence is just potential. Its ability to act is its power. By integrating your existing APIs and custom-built tools, you can transform bots.do agents from smart assistants into a fully-fledged digital workforce.

Ready to turn your business processes into autonomous agents? Explore the bots.do documentation and start building your first digital worker today.