Creating Agents

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

​

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.

​

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.

​

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.

​

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:

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

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

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

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

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