Make.com's New AI Agents (2026): The Complete Guide to Setup & Configuration
Most businesses struggle with fragmented automation tools that require constant switching between interfaces. Make.com's revolutionary 2026 AI agents combine brain, memory, tools and knowledge in one streamlined hexagon-shaped module - cutting configuration time by 70% while delivering smarter automation.
What Changed in Make.com's 2026 AI Agents?
If you've used Make.com's AI agents before February 2026, you'll immediately notice the revolutionary interface changes. Gone are the days of bouncing between multiple tabs to configure different components. The new system consolidates everything into one streamlined workflow.
The most visible change is the agent module itself - no longer a simple circle, but a distinctive hexagon shape that visually represents the interconnected nature of modern AI agents. As you build your agent, tools attach directly to this hexagon, creating a clear visual map of your automation's capabilities.
70% faster configuration: Early adopters report reducing setup time from 2-3 hours to just 30-45 minutes thanks to the unified interface. All components - from the LLM brain to memory settings - are now accessible within the scenario builder.
The 4 Essential Components of Every AI Agent
Understanding these four pillars is crucial before building your first agent. Each component serves a distinct purpose in creating intelligent, context-aware automation.
- The Brain: This is your LLM (GPT 5.2, Claude Opus 4.6, Gemini 3 Pro, etc.) that handles reasoning and decision making.
- System Prompt: Your agent's rulebook - instructions on how to process inputs, which tools to use, and how to generate outputs.
- Memory: Determines how much chat history context your agent retains (typically 10-50 previous messages).
- Tools: The "muscles" that let your agent take action (email, CRM updates, knowledge retrieval, etc.).
At 4:32 in the video tutorial, you'll see how these components work together in a law firm use case - where the agent retrieves case details, drafts emails, and schedules follow-ups automatically.
Step-by-Step Agent Setup Process
Creating your first agent is surprisingly simple with the new interface. Here's the streamlined workflow:
Step 1: Add the Agent Module
Search for "AI agent" in the module selector - the new hexagon-shaped module appears at the top. Unlike previous versions, there's only one agent type now.
Step 2: Configure the Brain
Select your AI provider (Make's default or connect your own API key for OpenAI, Anthropic, etc.). Pro tip: Start with a powerful model like GPT 5.2 during development, then test cheaper options later.
Step 3: Set Your System Prompt
This is your agent's instruction manual. Include: mission statement, behavior rules, tools list, and step-by-step task flows. The video shows a 7-part prompt structure at 7:15 that works for most use cases.
Step 4: Connect Input Sources
Define where your agent receives messages - Slack channels, web forms, WhatsApp, etc. The interface now lets you test inputs directly without leaving the scenario.
Configuring the Brain (LLM Selection)
Your agent's cognitive power comes from its LLM. Make.com offers two configuration paths:
Critical rule: Always start development with a more capable model than you think you'll need. This ensures tool integration and prompt effectiveness before optimizing for cost.
Option 1: Make's AI Providers
Simple dropdown selection of curated models. As of February 2026, this includes GPT 5.2, Claude Opus 4.6, and Gemini 3 Pro. Pricing is built into your Make plan.
Option 2: Your Own API Connections
For paid plans, you can connect directly to OpenAI, Anthropic, Google Gemini, etc. This gives access to specialized models but requires managing API costs separately.
At 9:45 in the video, you'll see how to connect an OpenAI account and select between GPT models based on your needs and budget.
Crafting the Perfect System Prompt
Your system prompt is the rulebook that makes your agent effective. The new interface provides a dedicated text area with syntax highlighting for better readability.
A well-structured prompt includes:
- Mission Statement: "You are a legal assistant AI for Mason Law Firm..."
- Behavior Rules: "Always verify facts against the knowledge base..."
- Tools List: "Available tools: 1) Case lookup, 2) Email drafter..."
- Task Flows: "When asked about a case: 1) Retrieve details 2) Summarize 3) Offer follow-up actions"
The video demonstrates a complete 7-part prompt structure at 7:15 that handles edge cases while maintaining natural conversation flow.
Memory Settings Explained
Memory is what separates basic bots from true AI agents. The new interface makes memory configuration straightforward:
Accessing Memory Settings: Go to Advanced Settings at the bottom of the agent configuration panel. Scroll down to "Memory Context".
How It Works: This setting determines how many previous messages your agent can reference. A setting of 10 means the agent remembers the last 10 exchanges in the conversation.
Pro Tip: For complex workflows like legal case management shown at 18:30, set memory to 30-50 messages. This prevents repetitive questions while maintaining context across multi-step processes.
Adding & Configuring Tools
Tools give your agent the ability to take action. The new interface shows attached tools visually beneath the agent hexagon.
Adding Tools: Hover the plus icon next to your agent and select "Add Tool". Choose from hundreds of pre-built integrations like Gmail, Google Sheets, or Slack.
Tool Configuration: Each tool needs:
- A clear name ("Send Follow-Up Email")
- Brief description under 240 characters
- Connection to your service account
- Field settings (let AI decide dynamic fields)
At 14:20 in the video, you'll see how the law firm agent dynamically populates email fields based on conversation context.
Working With Knowledge Bases
Make.com's built-in knowledge base handles document uploads directly in the interface - a huge improvement over previous external configurations.
Supported Files: txt, PDF, Word, CSV, MD, JSON (20MB max per file or 20 files max per agent)
Upload Process: Click the knowledge icon on your agent, upload files, and they're automatically vectorized for semantic search.
For larger needs: As shown at 22:10, integrate with Pinecone or Supabase for enterprise-scale knowledge. The video demonstrates calling a separate Pinecone scenario from within the agent.
Testing Agents Before Deployment
The new interface includes native testing - no more deploying to Slack just to debug your agent.
Testing Methods:
- Right-click the agent module → "Chat with Agent"
- Use the chat bubble icon in the scenario
Testing Strategy:
- Verify tool triggering (does it use the right tool for each task?)
- Check knowledge retrieval accuracy
- Test edge cases and unexpected inputs
At 17:45, the video shows comprehensive testing of the law firm agent - from case lookups to email drafting and calendar scheduling.
Moving to Production Deployment
Once testing is complete, connect your agent to real business channels:
Common Deployment Channels:
- Slack: Watch specific channels for mentions or DMs
- Web Forms: Process contact form submissions in real-time
- WhatsApp/Telegram: Customer-facing conversational interfaces
Monitoring Tips:
- Start with a limited rollout (one channel or team)
- Review agent decisions weekly for refinement
- Expand to more channels as confidence grows
The video concludes at 25:30 with the law firm agent fully deployed in Slack, handling case inquiries and scheduling autonomously.
Watch the Full Tutorial
For visual learners, the complete 20-minute video tutorial demonstrates every step from blank scenario to fully deployed agent. Pay special attention at 14:20 where we configure dynamic email tools and 22:10 for advanced Pinecone integration.
Key Takeaways
Make.com's 2026 AI agent update represents a quantum leap in automation accessibility. By consolidating all components into one interface and adding native testing, they've reduced the barrier to creating sophisticated business automation.
In summary: 1) Hexagon modules visually organize tools, 2) Memory settings maintain conversation context, 3) Built-in testing prevents production errors, and 4) Consolidated configuration cuts setup time by 70%. These agents can transform customer service, internal operations, and multi-step workflows when properly implemented.
Frequently Asked Questions
Common questions about Make.com AI agents
The four core components are: 1) The brain (LLM like GPT 5.2 or Claude Opus), 2) System prompt (rules and instructions), 3) Memory (chat history context), and 4) Tools (actions the agent can execute).
These components work together to create intelligent automation that understands context and takes appropriate actions. The brain handles reasoning, the prompt provides direction, memory maintains conversation flow, and tools enable real-world actions.
- Brain selection impacts reasoning quality and cost
- System prompts require careful crafting for best results
- Memory settings typically range from 10-50 messages depending on use case
The new interface consolidates everything into one scenario with hexagon-shaped agent modules. Previously, you had to configure agents in separate tabs, then connect them to scenarios.
Now all configuration happens directly in the scenario with built-in tools, knowledge uploads, and testing capabilities. This eliminates the back-and-forth between different configuration screens that slowed down development in previous versions.
- 70% faster setup reported by early users
- Visual tool attachment to hexagon modules
- Native testing without external deployments
Make.com's built-in knowledge base supports txt, PDF, Word docs, CSV, MD, and JSON files. There's a 20MB maximum per file or 20 files maximum per agent.
For many small business use cases (FAQs, product info, basic documentation), this built-in solution works perfectly. The system automatically vectorizes uploaded files for semantic search capabilities.
- Text extraction works best with clean PDFs and Word docs
- CSV files should have clear column headers
- For larger needs, integrate Pinecone or Supabase
Memory configuration is found in Advanced Settings. You specify how many previous messages the agent can reference for context (default is 10). For complex workflows, you might set this to 30-50 messages.
This allows the agent to maintain conversation continuity without repeating questions. In the law firm example from the video, a setting of 30 messages lets the agent reference multiple previous inquiries about a case.
- Higher settings increase context but also costs
- Start with 10, increase if conversation flow breaks
- Reset memory between unrelated conversations
Always start with a more capable model (like GPT 5.2) during development. Once your agent works perfectly, test with cheaper models until you find the sweet spot where quality remains but costs decrease.
This ensures you don't sacrifice performance while optimizing expenses. The video shows this process at 9:45, starting with GPT 5.2 then testing Claude Haiku for cost efficiency.
- Powerful models handle edge cases better during development
- Once stable, smaller models often suffice for routine tasks
- Monitor quality metrics when downgrading models
Make.com provides native testing within the scenario interface. Right-click the agent module and select 'Chat with Agent' to interact directly. Test all tools and edge cases before connecting to live channels like Slack or WhatsApp.
This prevents errors in production environments. The video demonstrates comprehensive testing at 17:45, verifying knowledge retrieval, tool triggering, and conversation flow.
- Test with realistic but varied inputs
- Verify all tools trigger appropriately
- Check knowledge retrieval accuracy
Common use cases include: 1) Customer service agents in Slack/WhatsApp, 2) CRM automation (updating records, scheduling follow-ups), 3) Knowledge retrieval for teams, 4) Automated email responses, and 5) Multi-step workflow automation combining multiple business tools.
The video focuses on a law firm use case where the agent handles case inquiries, drafts emails, and schedules meetings - demonstrating how one agent can replace multiple manual processes.
- 24/7 customer service with instant responses
- Automated data entry and CRM updates
- Internal knowledge base for employee questions
GrowwStacks specializes in building custom AI agent solutions on Make.com. We'll design agents tailored to your workflows, integrate your existing tools, and ensure smooth deployment.
Our team handles everything from system prompts to tool configuration and ongoing optimization. We've implemented agents for legal firms, eCommerce stores, and professional service providers - typically seeing 60-80% time savings on automated processes.
- Free consultation to map your automation opportunities
- Custom agent development for your specific needs
- Ongoing support and optimization services
Ready to Transform Your Business With AI Agents?
Manual processes are costing you hours every week that could be spent growing your business. Our Make.com experts will build custom AI agents that handle repetitive tasks automatically - just like the law firm example in this guide.