> ## Documentation Index
> Fetch the complete documentation index at: https://docs.beam.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Creating Agents

> Build AI agents in Beam using templates, chat-based generation, or custom configuration from scratch

Creating AI agents in Beam can be approached through multiple pathways, each designed for different use cases and skill levels.

<Info>
  **Quick Start**: New to Beam? Start with a template to understand core concepts, then progress to custom builds as your needs evolve.
</Info>

## Agent Creation Methods

### Template-Based Creation

**Best for**: Quick deployment, learning fundamentals, standard use cases

**Time to deploy**: 5-15 minutes

**Skill level**: Beginner to Intermediate

Templates provide pre-configured agents with **complete workflows already built and ready to deploy**. Each template includes a fully connected flow with nodes, tools, and configurations set up for specific use cases. You can immediately view the entire workflow structure, understand how it operates, and customize it to match your exact requirements. Perfect for common business scenarios like Customer Service, Sales Operations, and Data Extraction.

<iframe src="https://app.supademo.com/embed/cmgro9a7u0oymyzgy16a8wo39" frameborder="0" webkitallowfullscreen="true" mozallowfullscreen="true" allowfullscreen style={{width: "100%", height: "450px"}} />

<Steps>
  <Step title="Access Templates">
    Navigate to Agent Templates in the sidebar or use the search bar to find "Create a new agent from a template"
  </Step>

  <Step title="Select Your Template">
    Choose the template that matches your use case:

    * **Customer Service**: Email processing, reply generation, escalation routing
    * **Sales Operations**: Lead qualification, CRM updates, follow-up automation
    * **Data Extraction**: Document processing, structured output generation

    <Frame>
      <img src="https://mintcdn.com/beamai/_fu-3ODo4WjBAai0/02-building-agents/agent-fundamentals/creating-agents/Screenshot%202025-11-03%20at%2023.07.41.png?fit=max&auto=format&n=_fu-3ODo4WjBAai0&q=85&s=cae60e81b1cb9128075c38327a968f2f" alt="" width="2300" height="1268" data-path="02-building-agents/agent-fundamentals/creating-agents/Screenshot 2025-11-03 at 23.07.41.png" />
    </Frame>
  </Step>

  <Step title="Review Pre-Built Flow">
    Once created, you'll see the **complete workflow** with all nodes already connected and configured. Review:

    * Node sequence and connections
    * Tools assigned to each node
    * Variable mappings and data flow
    * Branch conditions and exit points

    <Frame>
      <img src="https://mintcdn.com/beamai/_fu-3ODo4WjBAai0/02-building-agents/agent-fundamentals/creating-agents/Screenshot%202025-11-03%20at%2023.24.08.png?fit=max&auto=format&n=_fu-3ODo4WjBAai0&q=85&s=12de7b9e61e9e63a3c03359b40cad0dd" alt="" width="2362" height="1410" data-path="02-building-agents/agent-fundamentals/creating-agents/Screenshot 2025-11-03 at 23.24.08.png" />
    </Frame>
  </Step>

  <Step title="Modify As Needed">
    Customize the template to match your specific requirements:

    * Adjust node objectives and tools
    * Modify prompts for your domain
    * Add or remove nodes
    * Reconfigure variable mappings

    For detailed flow modifications, see [Creating Flows](/02-building-agents/agent-fundamentals/flow-configuration/flow-configuration).
  </Step>
</Steps>

### Chat-Based Creation

**Best for**: Rapid prototyping, conversational workflows, quick iterations

**Time to deploy**: 10-20 minutes

**Skill level**: Beginner to Intermediate

Describe your automation need in natural language and Beam **automatically generates a complete flow** with connected nodes. You receive a working workflow that you can immediately view, test, and modify.

<iframe src="https://app.supademo.com/embed/cmgi7uzde081gkrn98r49v4nb" frameborder="0" webkitallowfullscreen="true" mozallowfullscreen="true" allowfullscreen style={{width: "100%", height: "450px"}} />

<Steps>
  <Step title="Describe Your Process">
    Type a description of what you want to automate. Be specific about:

    * Input data sources
    * Desired actions and transformations
    * Expected outputs
  </Step>

  <Step title="Review Generated Flow">
    Beam creates a complete agent workflow based on your description. Review:

    * Workflow nodes and connections
    * Tools automatically selected for each node
    * Variable configurations
    * Input/output schema
  </Step>

  <Step title="Refine and Test">
    Adjust the generated workflow as needed:

    * Modify node configurations
    * Adjust prompts and tool selections
    * Test with sample data

    For flow adjustments, refer to [Creating Flows](/02-building-agents/agent-fundamentals/flow-configuration/flow-configuration).
  </Step>
</Steps>

### Custom Agent Creation

**Best for**: Unique workflows, complex requirements, advanced users

**Time to deploy**: 30-90 minutes

**Skill level**: Intermediate to Advanced

Building from scratch offers complete control over agent configuration for precise business requirements. You start with an **empty canvas** and build your flow node by node.

<Steps>
  <Step title="Start with Empty Canvas">
    Click "Create custom agent" to open the visual flow builder with a blank workspace.
  </Step>

  <Step title="Add Your First Node">
    Create the first node in your workflow:

    1. Click the **+ icon** on the canvas
    2. Add an **objective** describing what the node should accomplish
    3. **Select the tool** for this node:
       * **Match**: Find existing tools that fit the objective
       * **Select from catalog**: Browse integration connectors
       * **Generate new**: Create a new Custom GPT tool

    **Node Data Access:**

    Every node you create has access to:

    * **`task_query`**: The trigger input that started the workflow
    * **Previous node outputs**: Via Linked variable fill method
    * **Memory context**: Historical data from past interactions
    * **Static values**: Configuration data defined in the workflow
  </Step>

  <Step title="Build Your Complete Flow">
    Continue adding nodes, branches, and configuring the workflow.

    **See [Creating Flows](/02-building-agents/agent-fundamentals/flow-configuration/flow-configuration) for detailed guidance on:**

    * Adding and connecting multiple nodes
    * Setting up branches and conditional logic
    * Configuring variable mappings
    * Defining merge points
    * Setting exit conditions
  </Step>
</Steps>

<Info>
  **Every Agent Has a Flow**: Regardless of creation method, all agents execute through workflows. Templates and chat-based agents provide pre-built flows you can modify, while custom agents start with an empty canvas where you build the flow from scratch.
</Info>

## Additional Configuration

After setting up your flow (via template, chat, or custom), configure these additional agent settings:

### Trigger Configuration

Define how your agent receives data and when it executes. Triggers determine what starts your agent's workflow and what data (`task_query`) gets passed to the first node.

**Adding Triggers:**

1. Navigate to your agent's **Triggers** tab
2. Click **"Add Trigger"** to open the trigger selection modal
3. Choose from available trigger types
4. Configure trigger-specific settings
5. Define the input schema that becomes `task_query`

**Available Trigger Types:**

<Accordion title="Integration-Based Triggers">
  Event-driven triggers that respond to actions in connected services:

  **New Pull Request** (GitHub)

  * Triggers when a new pull request is created
  * Use case: Automate PR review, testing, or notifications
  * Input: PR details, author, files changed, description

  **New Email Received** (Gmail)

  * Triggers when new email arrives in Gmail inbox
  * Use case: Auto-categorize emails, extract data, route to teams
  * Input: Email subject, body, sender, attachments

  **File or Folder Updated** (Google Drive)

  * Triggers when files or folders change in Google Drive
  * Use case: Process uploaded documents, sync data, notify teams
  * Input: File name, path, content, modification details

  **New Mail Received** (Outlook)

  * Triggers when new email arrives in Outlook
  * Use case: Email triage, data extraction, automated responses
  * Input: Email metadata, content, sender information

  **New Message Received** (Slack/Teams)

  * Triggers when new message posted in channels
  * Use case: Monitor conversations, extract action items, respond to requests
  * Input: Message content, channel, sender, timestamp
</Accordion>

<Accordion title="Schedule-Based Triggers">
  **Timer**

  * Runs agent on defined time intervals
  * Use case: Daily reports, periodic data syncs, scheduled analysis
  * Configuration: Set cron schedule or interval (hourly, daily, weekly)
  * Input: Timestamp, scheduled execution metadata
</Accordion>

<Accordion title="HTTP-Based Triggers">
  **Webhook**

  * Receives HTTP requests from external systems
  * Use case: Custom integrations, API endpoints, third-party notifications
  * Configuration: Beam provides webhook URL endpoint
  * Input: HTTP request body, headers, query parameters
  * Authentication: Optional API key or signature validation
</Accordion>

<Accordion title="Manual Triggers">
  **Manual Execution**

  * User manually starts the agent
  * Use case: On-demand processing, testing, ad-hoc operations
  * Input: User-provided form data or predefined test inputs
</Accordion>

**Trigger Configuration Best Practices:**

* **Define Clear Input Schema**: Specify exactly what data the trigger provides to `task_query`
* **Test Triggers**: Use test mode to verify trigger data before production deployment
* **Multiple Triggers**: Add multiple triggers to the same agent for different entry points
* **Authentication**: Configure authentication for webhook triggers to secure endpoints
* **Error Handling**: Set up error notifications if trigger fails or receives malformed data

**Example Use Cases:**

**Customer Support Agent:**

* Trigger: New Email Received (Gmail)
* Input: Email content, sender, subject
* Flow: Classify inquiry → Route to appropriate team → Auto-respond

**Code Review Agent:**

* Trigger: New Pull Request (GitHub)
* Input: PR details, changed files, description
* Flow: Analyze code → Run tests → Post review comments

**Daily Report Agent:**

* Trigger: Timer (Daily at 9 AM)
* Input: Current date, reporting parameters
* Flow: Fetch metrics → Generate report → Send via email

### Memory Files

Upload files to agent memory for use across workflow executions:

**When to Use Memory Files:**

* Reference documents the agent needs to consult
* Knowledge bases for context-aware responses
* Product catalogs, pricing sheets, documentation
* Templates or examples for generation tasks

**Adding Files to Memory:**

1. Navigate to agent Memory tab
2. Upload files (PDF, CSV, TXT, JSON, etc.)
3. Files become accessible to all nodes via "From Memory" variable fill method
4. Agent can search and retrieve relevant content during execution

## Next Steps

<CardGroup cols={2}>
  <Card title="Creating Flows" icon="diagram-project" href="/02-building-agents/agent-fundamentals/flow-configuration/flow-configuration">
    Deep dive into building workflows with nodes, branches, and logic
  </Card>

  <Card title="Selecting Tools" icon="wrench" href="/02-building-agents/agent-fundamentals/tools-integrations/tools-integrations">
    Choose and configure tools for your workflow nodes
  </Card>

  <Card title="Variable Fill" icon="arrow-right-arrow-left" href="/02-building-agents/agent-configuration/variables-state/variables-state">
    Configure how data flows through your workflow
  </Card>

  <Card title="Integrations" icon="plug" href="/02-building-agents/agent-configuration/integrations/integrations">
    Connect external services and APIs to your agents
  </Card>
</CardGroup>