Base44 Super Agents Tutorial: Build Your First AI Agent in 30 Minutes (2026)
Last Updated: March 26, 2026 · 28 min read
A complete, step-by-step walkthrough for building, testing, and deploying your first Base44 Super Agent — including the exact instructions, entity setup, and integration config you need to get a working agent live today.
You don't need to write code. You don't need a computer science degree. You don't need to know what Docker, SSH, or an API key is. You just need 30 minutes and this tutorial.
By the end of this guide, you'll have a fully functional AI agent running inside Base44 that can research leads, draft personalized outreach emails, and save everything to a structured database — automatically. No human in the loop.
This isn't a conceptual overview. This is the exact, click-by-click process for building a real Super Agent from scratch. Every step, every instruction template, every entity field, and every integration config is included. Follow along and you'll have a working agent before you finish your coffee.
Ready to follow along?
Create your free Base44 account and build your first agent while you read this tutorial.
What Is a Base44 Super Agent?
A Base44 Super Agent is a persistent, autonomous AI assistant that connects to your tools, monitors events, and takes real action across your workspace — all without you having to be involved.
A Super Agent can read your emails and respond to them. It can add rows to your CRM. It can send follow-up messages on WhatsApp. It can compile research from the web, organize it into structured data, and send you a summary on Slack every morning at 7 AM. It can monitor your Google Calendar and text you a daily briefing. It can follow up with leads who haven't responded in 3 days — using a different email angle each time.
It does all of this without you opening a laptop. Once you build and deploy it, the agent runs continuously in the background, 24 hours a day, 7 days a week.
Base44 launched Super Agents on March 11, 2026. The platform was acquired by Wix in June 2025, so you're building on infrastructure backed by one of the largest website and software companies in the world. This isn't a weekend project that might go offline.
How Is This Different from ChatGPT or Claude?
ChatGPT / Claude
Reactive — you ask, they answer. You close the window, they stop. Everything resets. Think of it as a consultant you have to call every time you need something.
Super Agent
Proactive — has its own schedule, memory, and tools. Doesn't wait for you to ask. Connected to Gmail, Slack, Google Calendar, CRMs. Like a full-time employee who comes in every day and just does the job.
Why Super Agents Are Worth Learning Right Now
The market is wide open
Super Agents launched two weeks ago. The number of people who know how to build them well is tiny. The number of businesses that want them is enormous and growing daily. Every week that passes, the gap between supply and demand gets wider.
Businesses already understand the value
You don't have to explain what AI is anymore. Every business owner has used ChatGPT. The pitch isn't 'AI is useful' — it's 'let me show you an AI that runs your lead follow-up while you sleep.' They get it immediately.
The barrier to entry is zero
You don't need to code. You don't need technical experience. The platform handles everything: hosting, databases, security, deployment. All you need is the ability to describe what a business needs and configure the agent to do it.
What You'll Build in This Tutorial
We're going to build a Lead Research and Outreach Agent — one of the highest-value agents you can sell to B2B businesses. Here's exactly what it does:
- 1Takes a list of company names or website URLs as input
- 2Automatically researches each company using web search
- 3Extracts key information: industry, company size, recent news, and likely pain points
- 4Drafts a personalized cold outreach email for each lead — referencing something specific about the company
- 5Saves everything to a structured database with status tracking
- 6Optionally sends the email automatically or queues it for review
Prerequisites (Almost None)
That's it. No software to install. No terminal to open. No dependencies to manage. No credit card required to start.
Set Up Your Base44 Account
Head to Base44 and sign up here to get started for free — no credit card required. You can register with your Google account or email address.
Once you're signed in, you'll land on the Base44 dashboard. Base44 gives you free credits to start building immediately.
Create Your Super Agent
Navigate to the Super Agent section from your dashboard homepage. Click to create a new Super Agent.
You'll be prompted to describe what you want your agent to do. Type something like:
"I want to build a lead research agent that takes a list of company names, researches each one automatically, finds their industry, size, recent news, and pain points, then writes a personalized cold outreach email for each lead."
Base44 will process this and generate an initial structure: instructions, suggested tools, and a basic configuration. This is your starting point — a first draft, not a finished product. Over the next steps, we'll refine every piece of this until it performs at a professional level.
Define Your Entities (Data Model)
Before your agent can do anything useful, it needs a place to store structured data. In Base44, these are called Entities — like tables in a spreadsheet, but with types, relationships, and the ability to be read and written by your agent automatically.
For our Lead Research Agent, create a Lead entity with these fields:
| Field Name | Type | Purpose |
|---|---|---|
| company_name | Text | The name of the target company |
| website | Text | Company URL for research |
| industry | Text | Auto-detected by the agent |
| company_size | Text | Approximate employee count or revenue band |
| headquarters | Text | Company location |
| recent_news | Long Text | Latest developments, funding, launches, etc. |
| pain_points | Long Text | Industry-specific challenges the company likely faces |
| key_contact | Text | Decision-maker name if found |
| contact_title | Text | Decision-maker's role |
| personalized_email | Long Text | The outreach email the agent drafts |
| email_subject_line | Text | Subject line for the outreach email |
| status | Select | Options: pending / researched / email_drafted / outreach_sent / replied |
| research_date | Date | When the research was conducted |
| notes | Long Text | Any additional context or flags |
Want to see this entity structure live? Follow along on Base44 — free to start.
Open Base44 →Optional: Add a Campaign Entity
If you plan to manage multiple outreach campaigns, add a Campaign entity. This allows your agent to tailor outreach based on which campaign a lead belongs to — and it's what separates a $1,500 agent from a $4,000 agent.
| Field | Purpose |
|---|---|
| campaign_name | Name of the outreach campaign |
| target_industry | What industry you're targeting |
| target_company_size | Size range of target companies |
| value_proposition | What you're offering this segment |
| email_template_notes | Guidance for email personalization |
Write Your Agent Instructions
This is the most important step in the entire build. Your instructions are the brain of your agent. Everything else — the entities, the tools, the automations — are just infrastructure. The instructions determine whether your agent produces amateur-hour output or client-ready results.
The Complete Instruction Template for a Lead Research Agent
Copy this and customize it for your specific use case:
Identity
You are Alex, a senior B2B lead research specialist with 8 years of experience in sales intelligence. You work for [Your Company Name]. You're known for finding specific, actionable insights about target companies — not generic information anyone could find in 10 seconds. Your research consistently leads to higher response rates because it demonstrates genuine understanding of each company's situation.
Primary Objective
Your primary job is to research target companies thoroughly and produce personalized cold outreach emails that get responses. Every piece of research you do should answer one question: "What specific thing about this company can I reference that proves I actually looked into them?"
Research Workflow
When given a company name and/or website URL, execute the following process in order: 1. Visit the company website and identify what they do, who they serve, and what their core product or service is. 2. Determine the company's industry classification and approximate size (use ranges: 1-10, 11-50, 51-200, 201-1000, 1000+). 3. Search for recent news from the past 90 days. Prioritize: funding rounds, product launches, leadership changes, expansions, partnerships, or public challenges. 4. Based on the industry and company size, identify 2-3 likely pain points — based on evidence, not assumptions. 5. If possible, identify a key decision-maker (CEO, VP of Sales, Head of Marketing, etc.) and their title. 6. Draft a personalized cold outreach email that: opens with a specific reference to the company, connects to a problem you can help solve, proposes a specific next step, is under 150 words, uses a professional but conversational tone, and includes a subject line under 8 words. 7. Save all findings to the Lead entity. Set status to "email_drafted."
Quality Standards
• Never use generic phrases like "I came across your company" or "I hope this email finds you well." • Every email must reference something specific and verifiable about that company. • Pain points must be evidence-based — cite industry, size, stage, or recent activity. • If you can't find sufficient information to write a quality email, set status to "needs_manual_review" and explain what's missing. A bad email is worse than no email.
Constraints
• Never fabricate information about a company. If unsure, say so. • Never claim to have met the recipient or reference a non-existent relationship. • Never include pricing in cold outreach — the goal is to start a conversation. • Never send emails automatically unless the user explicitly enables auto-send. Default is draft-and-save. • Keep all emails under 150 words.
Why These Instructions Work
Defined persona — the agent has a name, experience level, and reputation. It helps maintain consistent behavior and quality standards across dozens of leads.
Specified workflow — not just "do research" but a numbered sequence of specific steps. Consistent output quality regardless of the company being researched.
Quality gates — if the agent can't find enough information, it flags the lead for manual review instead of producing garbage output.
Explicit constraints — rules about what NOT to do. Without constraints, agents drift toward generic, safe output.
Upload Knowledge Files
The Knowledge tab is where you upload reference material your agent uses when making decisions and writing outputs.
Email Templates and Swipe Files
Create a markdown file with 3–5 example outreach emails that demonstrate the style and quality you want. Include examples for different scenarios: a company that just raised funding, one that recently launched a product, one in a specific industry you frequently target, and one where you found a specific pain point.
Industry Pain Point Reference
Create a document that maps common industries to their typical pain points. For example: • SaaS (51-200 employees): Scaling outbound sales, reducing churn, improving onboarding, managing technical debt • Marketing agencies (11-50 employees): Client retention, scope creep, hiring specialized talent, proving ROI • E-commerce ($1M–$10M revenue): Customer acquisition cost, inventory management, post-purchase engagement, paid ad saturation This grounds your agent's pain point analysis in real industry knowledge, not generic guesses.
Your Value Proposition
Create a document clearly explaining what you (or your client) offer, who it's for, and what results it delivers. The agent uses this to connect the prospect's pain points to a relevant solution in the outreach email.
Connect Your Tools and Integrations
Your agent needs tools to take action.
Essential Integrations
Recommended Integrations
Set Up Automations and Triggers
Automations are what make your agent truly autonomous. Without them, you have to manually prompt it every time. With them, it runs on its own.
Trigger 1: New Lead Added
EssentialWhen a new company name is added to the Lead entity with status 'pending,' automatically begin the research workflow. This means anyone — you, your client, or another system — can simply add a company name to the database, and the agent picks it up and starts working. No manual prompting required.
Trigger 2: Daily Summary
RecommendedEvery day at 8 AM, compile a summary of all leads researched in the last 24 hours and send it to Slack (or WhatsApp, or email). Include: number of leads researched, emails drafted, and any leads flagged for manual review. This gives your client a daily briefing without anyone asking for it — and demonstrates the agent is actively working.
Trigger 3: Follow-Up Reminder
OptionalIf a lead's status has been 'outreach_sent' for more than 3 days with no reply, automatically draft a follow-up email with a different angle and update the status to 'follow_up_drafted.' This turns your agent from a one-shot research tool into a persistent sales system. The follow-up is where most deals are won — and where most humans drop the ball.
Build a Simple UI
Your agent needs an interface. Describe these elements to Base44 in natural language, and the platform generates the UI automatically:
Input Section
- →Text input to paste company names or URLs (one per line)
- →Dropdown to select the campaign
- →'Research Leads' button to trigger the agent
Results Dashboard
- →Table with columns: company name, industry, size, status, research date
- →Color-coded status badges (gray=pending, blue=researched, green=email drafted, gold=replied)
- →Click-to-expand rows revealing full research output and drafted email
Email Review Panel
- →Drafted email with subject line in a clean preview format
- →'Approve & Send' and 'Edit' buttons
- →Research summary alongside the email so the user understands the context
Summary Stats
- →Total leads researched this week
- →Emails drafted and sent
- →Response rate (if tracking replies)
Designing for Clients
- Less is more — show the 4–5 most important columns. Let them expand for details.
- Make the primary action obvious — "Research Leads" and "Approve & Send" should be prominent.
- Use status badges — visual indicators communicate progress instantly without reading anything.
Test Your Agent (The Right Way)
Testing is where most beginners cut corners, and it's where the quality gap shows up. Don't skip this step.
Phase 1: Manual Testing (15 minutes)
Send your agent 5 real company names using a mix: a well-known company (tests filtering of results), a mid-size company in your target niche (tests usefulness), a small obscure company (tests handling of limited info), a company you know well (tests accuracy), and a company in a different industry (tests adaptability).
Phase 2: Evaluate Output Quality
For each company check: • Accuracy: Is the industry classification correct? Are recent news items real? • Specificity: Are pain points generic or specific and evidence-based? • Email quality: Does it reference something specific? Is it under 150 words? Would you actually send it? • Data integrity: Is every field populated? Is the status set correctly?
Phase 3: Iterate on Instructions
Based on evaluation, update the instructions: • Emails too generic → add more example emails to Knowledge files, strengthen Quality Standards • Pain points vague → expand the Industry Pain Point Reference • Agent fabricating info → add 'If you cannot verify this, mark as unconfirmed' to constraints • Inconsistent output format → add a rigid output template Most agents need 2–3 rounds of refinement before they're client-ready. This is normal.
Phase 4: Automation Testing
Once the core agent performs well manually, activate automations one at a time: 1. Enable 'new lead added' trigger. Add a test lead and verify the agent processes it. 2. Enable the daily summary. Wait for it to fire and verify format and content. 3. If using follow-up logic, set a test lead to 'outreach_sent' with a date 4 days ago and verify the follow-up drafts correctly. Test each automation individually — if you enable all at once and something breaks, you won't know which caused the issue.
Deploy and Share
Once your agent passes testing, it's ready to go live.
For Personal Use or Internal Teams
Your agent is already live and accessible from the Base44 dashboard. Share it by adding team members to your Base44 workspace. They can interact through the web interface or through connected messaging apps (WhatsApp, Telegram, Slack).
For Client Delivery
1. Help the client create a Base44 account (use your affiliate link). 2. Build inside their account or transfer your build. 3. Walk the client through connecting Gmail, CRM, and other integrations — do this on a call so they feel comfortable with the permissions. 4. Spend 15 minutes training them on the UI. Record the training call for their reference. 5. Agree on weekly check-ins for the first month. Review output together and make adjustments. This is where you demonstrate ongoing value that justifies a monthly retainer.
For Public Access
If you want to share your agent publicly (as a demo, lead magnet, or product), Base44 provides a shareable URL. You can also embed the agent into an existing website or link to it from social media.
The Instruction Template Library
The Lead Research Agent is just one build. Here are ready-to-use instruction frameworks for four more high-value agent types. Each follows the same structure: Identity, Primary Objective, Workflow, Output Format, Quality Standards, and Constraints.
Customer Support Agent Instructions
Identity
You are [Name], a patient and knowledgeable customer support specialist for [Company]. You've handled thousands of tickets and can resolve most issues without escalation.
Workflow
1. Categorize: billing, technical, account, general 2. Billing: check account status, explain charges, process simple changes 3. Technical: troubleshoot using knowledge base, attempt 2 solutions before escalating 4. Account: handle password resets, plan changes, profile updates 5. General: answer using FAQ or knowledge base 6. Log resolution in Tickets entity with category, resolution, and response time
Constraints
Never promise refunds without approval. Never share internal docs. Never guess — if unsure, escalate. Maximum response: 100 words unless a detailed walkthrough is needed.
Content Creation & SEO Agent Instructions
Identity
You are [Name], a content strategist specializing in [Industry/Niche]. You write content that ranks on Google and converts readers into leads.
Workflow
1. Research the target keyword: search volume, competition, top-ranking content 2. Analyze top 5 ranking pages: identify gaps and unanswered questions 3. Create a detailed content brief: keyword, word count, outline, unique angle 4. Write the full article 5. Optimize: keyword in title, H1, H2s, intro, and meta description 6. Save to Content entity with status tracking
Quality Standards
Content must be longer and more comprehensive than the current #1 result. Include specific examples, data points, or case studies. No generic advice.
Appointment Booking Agent Instructions
Identity
You are [Name], a scheduling coordinator for [Business]. You're organized, responsive, and make booking an appointment feel effortless for the customer.
Workflow
1. Check Google Calendar for available slots 2. Offer 3 time options matching the customer's preference 3. Once confirmed, create the calendar event and send confirmation via email and WhatsApp 4. 24 hours before: send a reminder with meeting link and prep materials 5. 1 hour before: send a final reminder 6. After: send a follow-up with next steps or feedback request
Constraints
Never double-book. Always offer rescheduling before canceling. If no slots match preference, propose the 2 closest alternatives.
Financial Reporting Agent Instructions
Identity
You are [Name], a financial analyst who turns raw numbers into clear, actionable summaries for business owners. You don't overwhelm with data — you highlight what matters.
Workflow
1. Pull key metrics: revenue, expenses, outstanding invoices, cash position 2. Compare to the same period last week and last month 3. Flag anomalies: unusual expenses, revenue changes, overdue invoices 4. Compile into a 3–5 bullet point summary 5. Send via WhatsApp or Slack at the scheduled time
Quality Standards
Every summary must be scannable in under 2 minutes. Lead with the single most important insight. Use simple language — no accounting jargon. Always include the cash position.
5 More Agent Builds You Can Do Today
Now that you understand the build process, here are five more agent ideas you can build using the exact same steps. Each one is a sellable product.
Get all 12 agent templates free — complete instruction sets, entity schemas, and knowledge file frameworks included.
Claim Templates →Competitor Monitoring Agent
$1,000–$2,500/moTracks competitor websites, social media, and news. Sends weekly summaries of competitor activity: new product launches, pricing changes, hiring trends, marketing campaigns.
Review & Reputation Management Agent
Valuable for local bizMonitors Google Reviews, Yelp, and social media mentions. Drafts responses to negative reviews. Alerts the business owner immediately when a 1-star review is posted. Compiles monthly sentiment reports.
Employee Onboarding Agent
$1,500–$3,000/moActs as an internal knowledge base that new hires can ask questions to. Knows the company handbook, policies, tool logins, and procedures. Reduces onboarding time in half.
Invoice & Payment Follow-Up Agent
Freelancers love thisMonitors outstanding invoices. Sends polite payment reminders at 7, 14, and 30 days overdue. Escalates to the business owner at 45 days. Tracks payment status and compiles aging reports.
Social Media Engagement Agent
$1,000–$2,500/moMonitors comments and DMs across social platforms. Responds to common questions using approved messaging. Flags potential leads based on keywords. Forwards high-priority messages to the business owner.
How to Sell This Agent to Clients
Setup Fee
$1,000–$3,000 one-time
Building, configuring, testing, deploying
Monthly Retainer
$500–$2,000/month
Maintenance, updates, optimization, credits
Want the exact client scripts and pricing templates? They're included free when you sign up through this link.
The Demo Framework
Don't pitch features. Demo outcomes.
Troubleshooting Common Issues
"My agent's output is too generic"
Cause: Instructions aren't specific enough. The agent defaults to safe, generic responses when it doesn't have clear direction.
Fix: Add more specific examples to your Knowledge files. Strengthen the Quality Standards section. Add explicit rules like 'every email must reference a specific, verifiable fact about the company.'
"My agent is making things up"
Cause: The agent doesn't have enough real data and fills gaps with fabricated information.
Fix: Add to constraints: 'If you cannot find verifiable information, set status to needs_manual_review and list what's missing. Never fabricate or assume information.' Also ensure web search is properly enabled as a tool.
"My automations aren't firing"
Cause: Usually a trigger misconfiguration or a permissions issue with a connected tool.
Fix: Check the trigger condition uses exact field names and status values. Verify all connected tools are still authorized — OAuth tokens can expire. Test the trigger with a simple case before adding complexity.
"I'm burning through credits too fast"
Cause: Automations running too frequently, or the agent is doing too much processing per task.
Fix: Reduce trigger frequency (hourly instead of every 5 minutes). Simplify the agent's workflow. Review which actions consume integration credits and optimize accordingly.
"The agent's tone doesn't match what I want"
Cause: Insufficient tone guidance in instructions.
Fix: Add 2–3 example interactions that demonstrate the exact tone you want. Be explicit: 'Use conversational language. Write at an 8th-grade reading level. Avoid formal greetings like Dear Sir/Madam. Use the recipient's first name.'
Frequently Asked Questions
Continue Reading
Get the Full Course Free
This tutorial covered one agent build from start to finish. The complete Super Agent Lab course covers 12 agent templates, client acquisition scripts, pricing frameworks, and everything you need to build a real income with Base44 Super Agents.
Updated regularly as Base44 releases new features. Last updated March 26, 2026.
