Unlock No-Code AI Agents: Claude + Make.com

In today's fast-paced business environment, AI agents can transform manual processes into intelligent automations. By combining Claude AI's advanced reasoning and tool-calling capabilities with Make.com's (formerly Integromat) visual workflow builder, you can create custom agents that handle multi-step tasks autonomously. This guide walks you through building a Sales Lead Qualifier Agent—a practical example that qualifies leads, researches companies, updates CRMs, and triggers notifications—all without writing code.

Whether you're in sales, marketing, or operations, this no-code mastery will save hours and boost efficiency.

Why Claude + Make.com for AI Agents?

• Claude's Strengths: Claude 3.5 Sonnet excels at tool use, reasoning, and handling complex instructions via the Anthropic API. It supports parallel tool calls and structured outputs, ideal for agentic flows.
• Make.com's Power: Drag-and-drop modules for triggers, HTTP API calls, iterators, routers, and integrations (e.g., HubSpot, Gmail, Google Search). Perfect for looping agent conversations.
• No-Code Advantage: Business users build enterprise-grade agents in minutes; developers extend with custom logic.
• Claude-Specific Edge: Unlike generic LLMs, Claude's constitutional AI ensures safer, more reliable outputs for business use.

Compared to Zapier (limited loops) or n8n (code-heavy), Make.com + Claude offers the best balance for scalable agents.

Prerequisites

• Make.com Account: Free tier works for testing; upgrade for production (starts at $9/mo).
• Anthropic API Key: Sign up at console.anthropic.com, generate a key (costs ~$3/million input tokens for Sonnet).
• Optional Integrations: HubSpot/Salesforce for CRM, Gmail/Slack for notifications, Google Custom Search API for research.
• Basic Familiarity: No coding needed, but understanding webhooks helps.

Step 1: Set Up Your Make.com Scenario

1. Log in to Make.com and click Create a new scenario.
2. Add a trigger module: Use Webhook > Custom webhook for testing (later connect to forms like Typeform).
   • Copy the webhook URL.
3. Test the webhook: Use a tool like webhook.site or Postman to POST sample lead data:

{
  "lead": {
    "name": "John Doe",
    "email": "john@example.com",
    "company": "Tech Startup",
    "message": "Interested in your API product."
  }
}

Click Run once to confirm data flows in.

Step 2: Initialize Claude Agent with Tools

Add an HTTP > Make a request module to call Claude API. Claude's tool-calling turns it into an agent "brain".

Configure:

• URL: https://api.anthropic.com/v1/messages
• Method: POST
• Headers:

  {
    "x-api-key": "YOUR_ANTHROPIC_API_KEY",
    "anthropic-version": "2023-06-01",
    "content-type": "application/json"
  }
  

• Body (JSON):

{
  "model": "claude-3-5-sonnet-20240620",
  "max_tokens": 1024,
  "messages": [
    {
      "role": "user",
      "content": [
        {
          "type": "text",
          "text": "Qualify this sales lead as hot/warm/cold. Research company if needed. Lead: {{1.lead}}. Tools available: research_company, update_crm, send_notification."
        }
      ]
    }
  ],
  "tools": [
    {
      "name": "research_company",
      "description": "Search for company info (revenue, size, news).",
      "input_schema": {
        "type": "object",
        "properties": {
          "company_name": {"type": "string"}
        }
      }
    },
    {
      "name": "update_crm",
      "description": "Add qualified lead to HubSpot.",
      "input_schema": {
        "type": "object",
        "properties": {
          "lead_data": {"type": "object"},
          "score": {"type": "string"}
        }
      }
    },
    {
      "name": "send_notification",
      "description": "Notify sales team via Slack.",
      "input_schema": {
        "type": "object",
        "properties": {
          "message": {"type": "string"}
        }
      }
    }
  ]
}

Replace {{1.lead}} with Make's variable from webhook.

Claude responds with tool_calls or final content.

Step 3: Parse Claude's Response and Handle Tools

Add a JSON > Parse JSON module:

• Data: {{2.body}}

Then, a Router module to branch:

• Filter 1: Final Answer (no tools): {{3.content[1].type}} equals text → End scenario or log.
• Filter 2: Tool Calls → Proceed to iterator.

Add Iterator for parallel tools:

• Array: {{3.content[1].tool_calls}} (handles multiple calls).

For each iteration, add Set Variable to prepare tool input.

Step 4: Execute Tools Dynamically

Use a Router after Iterator for tool type:

• Route 1: research_company → Google Search or SerpAPI module.

  • Query: {{iterator.company_name}} revenue funding
  • Output: Company summary.

• Route 2: update_crm → HubSpot > Create a Contact.

  • Map fields from {{iterator.lead_data}}.

• Route 3: send_notification → Slack > Send a Message.

  • Text: Hot lead: {{iterator.message}}

Aggregate results with Array Aggregator:

• Source: Tool outputs.
• Target: observations array.

Step 5: Loop Back to Claude (Agent Loop)

After aggregator, loop back to Claude HTTP module:

Update messages array:

{
  "model": "claude-3-5-sonnet-20240620",
  "max_tokens": 1024,
  "messages": [
    // Previous messages...
    {
      "role": "user",
      "content": "Previous tool observations: {{5.observations}}"
    },
    {
      "role": "assistant",
      "content": {{previous Claude response}}
    }
  ],
  "tools": [...] // Same tools
}

Use Repeater or error handler for 3-5 max iterations to prevent loops.

Step 6: Final Actions and Error Handling

After loop (when no tools called):

• Router based on Claude's final score:
  • Hot: HubSpot + Slack.
  • Warm: Email nurture sequence via Mailchimp.
  • Cold: Archive.

Add Tools > Sleep for rate limits, Error Handler routes for API failures.

Real-World Example: Lead Qualifier in Action

Input Webhook:

{"lead":{"name":"Jane Smith","email":"jane@acme.com","company":"Acme Inc","budget":"$50k","needs":"AI automation"}}

Claude Iteration 1:

• Calls research_company("Acme Inc") → Discovers $100M revenue, growing.

Iteration 2:

• Scores "Hot", calls update_crm and send_notification.

Output: Lead in HubSpot, Slack ping: "Hot lead from Acme Inc - $50k budget!".

This handles 100s of leads/day scalably.

Best Practices for Claude Agents in Make.com

• Prompt Engineering: Use XML tags for structure: <lead>{{lead}}</lead><instructions>Reason step-by-step.</instructions>.
• Token Limits: Monitor with usage in response; use Haiku for cheap research.
• Security: Store API keys in Make's Connection store.
• Testing: Use Make's Run history and Data store for logging.
• Scaling: Blueprints (export/import scenarios), teams collab.
• Cost Optimization: Batch tools, stop on final answer.
• Advanced: Multi-agent (Claude for reasoning + Haiku for speed), MCP servers for custom tools.

Troubleshooting:

Industry Playbooks

• Sales: Lead scoring → CRM sync.
• HR: Resume screening → Calendly booking.
• Support: Ticket triage → Zendesk assign.

Extend with Claude Code for hybrid flows.

Conclusion

You've now mastered no-code AI agents! Export this scenario as a template in Make.com and adapt for your needs. Claude's precision + Make's flexibility = unstoppable automation. Start building—your first agent takes <30 mins.

Next Steps:

• Try Opus for complex reasoning.
• Integrate MCP for Claude-native tools.
• Share your agents in comments!

(Word count: 1428)