Introduction to AgentRun AI
Artificial Intelligence is no longer a futuristic concept reserved for data scientists and software engineers. With the rise of accessible AI tools, anyone with a clear goal can now harness the power of intelligent automation. AgentRun AI is a platform designed to democratize the creation of AI agents. Whether you are a business owner looking to automate customer support, a marketer needing to scrape and summarize data, or a developer wanting to prototype a complex workflow, AgentRun AI provides a visual playground to build and deploy AI agents without writing extensive code.
At its core, AgentRun AI allows you to define a task, equip an AI agent with the necessary tools and instructions, and then set it loose to complete the work. The platform abstracts away the complexities of large language models (LLMs), API integrations, and server management, presenting a clean, drag-and-drop interface. This tutorial will guide you through the entire process, from your first login to deploying your first functional agent.
The official website is located at https://agentrun.ai/. We recommend creating a free account before proceeding with the steps below.
Getting Started with AgentRun AI
Creating Your Account and First Project
To begin, navigate to the AgentRun AI website and click on the “Sign Up” or “Get Started” button. You can typically register using an email address or a Google/GitHub account for a faster onboarding process. Once you have verified your email and logged in, you will be greeted by the main dashboard.
The dashboard is your command center. It displays your existing projects, available templates, and recent activity. To start fresh, look for a button labeled “New Agent” or “Create Project”. Clicking this will prompt you to name your agent (e.g., “Customer Support Bot” or “Research Assistant”). After naming it, you will be taken to the main workspace: the Visual Agent Builder.
Understanding the Interface
The Visual Agent Builder is the heart of AgentRun AI. It is a canvas where you will construct your agent’s logic. The interface is typically divided into three main areas:
- The Component Palette (Left Sidebar): This contains all the pre-built blocks you can use, such as “Input,” “LLM Call,” “API Request,” “Condition,” “Loop,” and “Output.”
- The Canvas (Center): This is your workspace. You drag components from the palette onto the canvas and connect them with arrows to define the flow of data.
- The Configuration Panel (Right Sidebar): When you click on a component on the canvas, this panel updates to show its settings. Here you can enter prompts, set API keys, define variables, and adjust parameters.
Take a few minutes to click around and familiarize yourself with the layout. Do not worry about breaking anything; you can always delete components or start a new project.
Key Features of AgentRun AI
Understanding the features available to you will help you design more powerful agents. Below are the core components you will use most frequently.
Visual Agent Builder
This is the flagship feature. Instead of writing code, you build a flowchart. You drag a “Trigger” block to start the agent, add an “LLM” block to ask a question, and connect it to a “Send Email” block to automate a response. This visual approach makes it easy to see the logic of your agent at a glance and debug issues by tracing the flow.
Pre-built Agent Templates
If you are unsure where to start, AgentRun AI offers a library of templates. These are pre-configured agents designed for common tasks such as:
- Content Summarizer: Input a URL or text, and the agent summarizes it.
- Lead Scraper: Extracts contact information from a website.
- Social Media Poster: Generates and posts content to platforms like Twitter or LinkedIn.
- Customer Support Triage: Classifies incoming messages and routes them to the right department.
You can use these templates as a starting point and modify them to fit your specific needs. This is the fastest way to learn how the blocks work together.
Workflow Automation
AgentRun AI is not just for one-off tasks. You can set your agents to run on a schedule (e.g., every hour, daily) or trigger them based on events (e.g., when a new email arrives, when a webhook is called). This transforms your agent into a fully autonomous worker that operates in the background.
Integration with External APIs
An AI agent is only as useful as the data it can access. AgentRun AI provides dedicated blocks for connecting to popular services like:
- Google Sheets (read and write data)
- Slack (send and receive messages)
- Notion (search and update databases)
- OpenAI / Anthropic (for the underlying LLM)
- Custom HTTP Requests (for any REST API)
This allows your agent to pull data from your CRM, update a spreadsheet with results, or send a notification to your team.
Deployment Options
Once your agent is built and tested, you need to make it available. AgentRun AI offers two primary deployment methods:
- Cloud Deployment: AgentRun hosts your agent on their servers. You get a public API endpoint or a webhook URL that you can call from your website or app.
- Local Deployment: For advanced users who need data privacy or low latency, you can export your agent configuration and run it on your own hardware using a provided runtime.
How to Use AgentRun AI: A Step-by-Step Guide
Let us build a practical agent together. We will create a “News Summarizer and Reporter” agent. This agent will take a topic, search the web for the latest news, summarize the top three articles, and then send the summary to your email.
Step 1: Set Up the Trigger
Start by dragging a “Manual Trigger” block onto the canvas. This means the agent will start when you click a “Run” button. In the configuration panel on the right, you can add a text input field. Label it “Topic” and set the type to “Text.” This is where you will type the news topic (e.g., “Artificial Intelligence in Healthcare”).
Step 2: Add a Web Search Block
From the palette, find the “Web Search” block (this may be labeled “Search” or “Google Search”). Drag it onto the canvas and connect the output of the “Manual Trigger” to the input of the “Web Search” block. In the configuration panel, you will need to provide an API key for a search engine (like SerpAPI or Bing Search). AgentRun AI usually provides a default option or instructions on how to get a free key. Set the “Query” field to use the variable from your trigger: {{Trigger.Topic}}. Set the number of results to 3.
Step 3: Extract Content from Articles
Search results give you URLs and snippets, but you need the full text for a good summary. Drag a “Fetch URL” or “HTTP Request” block onto the canvas. Connect the output of the “Web Search” block to this new block. You will need to configure this block to loop through the search results. Look for a setting called “Iterate over list” or “For each.” Configure it to take the URL from each search result and fetch the HTML content. The block will automatically extract the main text from the webpage.
Step 4: Summarize with an LLM
Now we need the AI brain. Drag an “LLM Call” block onto the canvas. Connect the output of the “Fetch URL” block to the LLM block. In the configuration panel, select your preferred model (e.g., GPT-4 or Claude). In the “System Prompt” field, enter a clear instruction:
“You are a professional news summarizer. You will receive the full text of a news article. Summarize it in 2-3 sentences, focusing on the key points. Keep the summary objective and factual.”
In the “User Prompt” field, use the variable that contains the extracted article text. It will look something like: “Please summarize the following article: {{FetchURL.Output}}”. Set the “Output” variable name to “Summary.”
Step 5: Compile the Final Report
You now have three summaries. To combine them into a single email, use a “Text Aggregator” or “Code Block” (with JavaScript or Python). This block will take the list of summaries and format them into a readable report. For example, you can use a simple JavaScript block with the following code:
return “Here is your news summary for ” + Trigger.Topic + “:nn” + LLMCall.Summaries.join(“nn—nn”);
Name the output of this block “FinalReport.”
Step 6: Send the Email
Finally, drag a “Send Email” block onto the canvas. Connect the output of the “Text Aggregator” to this block. Configure the recipient email address (you can use a variable or enter your own email). Set the subject to “News Summary: {{Trigger.Topic}}”. Set the body to {{TextAggregator.FinalReport}}. You will need to connect your email service (like Gmail or SMTP) in the settings. AgentRun AI will guide you through the OAuth or API key setup.
Step 7: Test and Deploy
Click the “Save” button in the top right corner. Then click “Run”. A sidebar will appear asking for the “Topic.” Type “Climate Change” and click “Submit.” Watch the agent flow through each block in real-time. If any block shows an error, click on it to see the error message and fix the configuration. Once the agent runs successfully, you will receive an email with your news summary.
To deploy, go to the “Deploy” tab. Choose “Cloud” and click “Deploy Agent.” You will receive a webhook URL. You can now call this URL from your website or set a schedule (e.g., run every day at 8 AM) to automate the process entirely.
Tips for Building Effective Agents with AgentRun AI
To get the most out of the platform, keep these best practices in mind.
Start Simple and Iterate
Do not try to build a complex agent on your first try. Start with a simple two-block agent (Input -> LLM -> Output) to understand how data flows. Once you are comfortable, add one block at a time. Test the agent after each addition. This makes debugging much easier.
Use Clear and Specific Prompts
The LLM block is the most powerful part of your agent, but it needs clear instructions. Avoid vague prompts like “Summarize this.” Instead, specify the format, length, and style. For example: “Summarize the following text in exactly three bullet points. Use simple language suitable for a 10-year-old.” The more context you give, the better the output.
Leverage Variables and Naming
When you connect blocks, data is passed via variables. Get into the habit of giving your blocks and outputs meaningful names. Instead of leaving the default “LLM_Output_1,” rename it to “Article_Summary.” This makes your workflow readable and prevents confusion when you have multiple LLM blocks in the same project.
Handle Errors Gracefully
External APIs can fail. Websites can be down. AgentRun AI allows you to add “Error Handling” paths. For example, if the “Fetch URL” block fails, you can route the flow to a “Log Error” block or a “Send Notification” block to alert you. This makes your agent robust and professional.
Use Templates as Learning Tools
When you are stuck, explore the template library. Open a template that does something similar to what you want. Do not just run it; study how the blocks are connected. Look at the prompts used in the LLM blocks. This is one of the fastest ways to learn advanced techniques like chaining multiple LLM calls or handling complex data transformations.
Monitor Usage and Costs
Every time your agent runs an LLM call or an API request, it consumes credits or tokens. AgentRun AI provides a usage dashboard. Keep an eye on it, especially if you set your agent to run on a schedule. Optimize your prompts to use fewer tokens (e.g., ask for shorter outputs) and cache results when possible to reduce costs.
Security First
If your agent handles sensitive data, consider using the Local Deployment option. Never hard-code API keys or passwords directly in prompts. Use the built-in “Secrets” or “Environment Variables” feature provided by AgentRun AI to store sensitive information securely.
By following this guide, you are now equipped to build, test, and deploy your own AI agents. The only limit is your imagination. Start with a small problem you face daily—like summarizing emails or tracking prices—and automate it with AgentRun AI.
