Welcome to AgentRunner beta. In this short video, I'll show you a quick example of how you can start using our AI agent-building software.
Getting Started with AgentRunner: Your Guide to Building AI Agents
Welcome to AgentRunner, a no-code platform designed to empower developers, AI engineers, product managers, and CTOs in building and managing AI agents with unparalleled efficiency. This guide will walk you through the essential steps to get started with AgentRunner, from connecting your AI provider to deploying your agents in a production environment. AgentRunner is engineered to provide the best in engineering platform capabilities, offering effective AI agent management, version control, detailed logs, and over 100 integrations with other software, with the integration list continuously growing. Best of all, AgentRunner is currently free to use during our beta phase!
Connecting Your AI Provider
The first step to unlocking the potential of AgentRunner is connecting your preferred AI models. This integration allows you to leverage the power of AI within your agent workflows.
Steps to Connect Your AI Provider
Navigate to the Apps Section: In the AgentRunner interface, find and click on the "Apps" section. This area is your gateway to connecting various AI providers.
Select Your AI Provider: AgentRunner currently supports OpenAI, Gemini, and Anthropic. Choose the AI provider you wish to connect.
Authenticate with API Keys: To establish a connection, you will need to provide the necessary API keys. These keys are essential for AgentRunner to securely access and utilize the AI models.
OpenAI: Obtain your API key from the OpenAI platform.
Gemini: Retrieve your API key from the Gemini AI platform.
Anthropic: Acquire your API key from the Anthropic platform.
Complete the Connection: Once you've entered your API keys, follow the on-screen prompts to finalize the connection. AgentRunner will verify the credentials and establish a secure link with your AI provider. Note: AgentRunner does not limit the number of AI provider applications you can connect, allowing you to harness the capabilities of multiple AI models within your projects.
Once you have successfully connected to your AI provider, you are ready to start building your AI agents on AgentRunner. This seamless integration sets the stage for creating innovative and scalable AI agents tailored to your specific needs.
Organizing Your Agents into Projects
With AgentRunner, you can efficiently organize your agents into projects, ensuring a structured and manageable environment for your AI initiatives. Projects serve as containers for your agents, allowing you to group them logically and control access for your team members.
Creating and Managing Projects
Create a New Project: Initiate a new project with a descriptive name that reflects its purpose. Provide a detailed description to outline the project's goals and scope.
Add Agents to the Project: Populate your project with AI agents, each designed to perform specific tasks or functions.
Utilize Folders for Organization: Within each project, create folders to further categorize your agents. This hierarchical structure enables you to maintain a clean and organized workspace.
Manage Team Access: For each folder, you can specify which team members have access. This feature allows you to create private folders accessible only to select individuals or open folders for broader team collaboration.
Understanding Access Roles: AgentRunner offers three distinct roles for team members:
Owner: Has full control over the project and its agents.
Developer: Can create and modify agents within the project.
Operator: Can run agents but cannot make changes to their configuration.
There is no limit of projects to be created, and there are no limits of how many agents you create.
Note: AgentRunner does not currently support moving agents between projects. Ensure that you place your agents in the appropriate project from the outset.
Editing and Running Your First AI Agent
The AgentRunner editor is where you bring your AI agents to life. This intuitive interface allows you to define the workflow of your agent, manage inputs and outputs, and test its performance.
Navigating the Agent Editor
Access the Agent Editor: Click on the AgentRunner icon associated with your agent to open the editor.
Define Inputs: Specify the input fields required for your agent to function. Give each input field a recognizable name to ensure clarity. AgentRunner supports text as the input type.
Add Nodes to Your Agent: Construct your agents by adding nodes. The available node types include:
Input: Defines the data that will be fed into the agent.
Output: Specifies the final result or data produced by the agent. An output node can include unlimited number of keys.
Text Merger: Combines multiple text inputs into a single output.
Conditional Check: Allows you to create branching logic based on specified conditions.
AI Nodes: Connects to the OpenAI, Gemini or Anthropic models to generate text or perform other AI tasks.
Apps: You can also integrate with over 100 external applications via API, and use other agents within the same project as nodes.
Configuring an AI Node
Add a Prompt: In the AI node, add your prompt & variables in the system or the user section.
Manage Controls: The controls tab allows you to manage the various parameters available for that AI provider such as next token, top p, top k, and the model selector.
Test Your Prompt: Utilize the "Test" tab to experiment with your prompt and variables. Enter example values to see the expected output. Iterate on your prompt until you achieve satisfactory results.
Running the Agent
Connect Variables: Ensure that all variables are correctly connected to the prompt and output nodes. Outputs from one node can be used as inputs for multiple subsequent nodes. Data is automatically passed between nodes in the format of the previous passing node's output.
Run the Agent: Click the "Run Agent" button to execute the workflow.
Enter Variables: On the left side of the editor, enter the values for your input variables.
View Results: The results of the workflow will be displayed on the right side of the editor. You can view the results as objects or as text.
Inspect Logs: Click the "View Details" button to access the logs for each step in the agent. The logs provide valuable information for debugging and troubleshooting. The logs include: an ID, timestamp, version, who it was run by, each API call, which API key has called it, how long it was running, and whether it was run successfully or is still running. For each step, you can see the inputs and outputs.
Creating a New Version of Your Agent
AgentRunner's version control system allows you to iterate on your agents without disrupting existing deployments. This feature is particularly useful for experimenting with new ideas and refining your agent's performance.
Steps to Create a New Version
Initiate a New Version: Create a new version of your agent. All modifications made will be specific to this new version.
Add Additional Nodes: Enhance your agent by adding additional nodes to the agent.
Connect Nodes: Connect the new nodes to the existing agent, utilizing outputs from previous nodes as inputs for subsequent nodes.
Run the New Version: Execute the new version of the agent to test its functionality.
Compare Results: Compare the results of the new version with the previous version to assess the impact of your changes.
Analyze Logs: Examine the logs to understand how each step of the agent's workflow was executed in the new version.
Each version is managed individually. To revert to a different version, simply select the desired version, and the editor will load its configuration. Each version can be run individually.
Calling Your Agent Through API
AgentRunner enables you to integrate your AI agents into other applications and systems via API calls. This feature allows you to leverage the power of your agents in a wide range of scenarios.
Generating API Keys
Navigate to Your Projects: Return to the projects dashboard.
Generate API Keys: Generate different API keys for each project. API keys are assigned to the project level. With an API key, you can only access nodes or agents within that project.
Secure Your API Keys: Protect your API keys as they are essential for accessing your agents. Only workspace owners can access these API keys. We recommend to turn on the two-factor authentication.
Making API Calls
Find the API Call Example: In each agent, you can find an example API call in cURL, including the variables and the version name.
Customize the API Call: Modify the API call to suit your specific needs, including specifying the input variables and the desired version of the agent.
Execute the API Call: Use the API key to authenticate the request and execute the API call.
Process the Response: Parse the response from the API call to extract the desired results.
Getting Help
AgentRunner provides various resources to assist you in your agent-building journey.
Accessing Help Resources
Click the Help Button: Click the help button in the left-hand menu to access our knowledge base articles.
Browse the Knowledge Base: Explore our comprehensive knowledge base for articles and tutorials on various aspects of AgentRunner.
Engage with the AI Chatbot: Ask our AI chatbot any questions. It is operated by our sister company called SAAS First, running on AgentRunner.
