Agents

​

What are Agents?

​

Core Components

​

How Agents Work

​

Dataset Processing

• Column assignment: Each column can have its own processing instructions
• Individual processing: Click research button on any cell to queue it
• Batch operations: Process entire columns with “Run All” or “Run Missing”
• Background processing: All work happens asynchronously without blocking workflow
• Evidence tracking: AI provides detailed reasoning, sources, and confidence indicators

​

Interactive Chat

• Understand datasets, tools, and ongoing projects
• Use connected tools during conversations
• Help with data analysis, planning, and decision-making
• Transparent tool usage with real-time execution

​

Creating Agents

1. Name and description
2. Select AI model and configure API key
3. Write system prompt defining role and behavior
4. Select relevant tools from workspace library
5. Test in chat interface before deploying

​

Management

​

Best Practices

​

Common Use Cases

​

Next Steps

​

Core Components

​

How Agents Work

​

Dataset Processing

• Assign agents to individual columns in your datasets
• Each column can have its own processing instructions
• Agents understand context from other columns to inform their work
• Process data individually or in batches across entire columns

• Individual Processing: Click the research button on any cell to queue it for processing
• Batch Operations: Process entire columns with “Run All” or “Run Missing” commands
• Background Processing: All work happens asynchronously without blocking your workflow
• Status Tracking: Monitor queued, processing, completed, and error states in real-time

• Agents provide detailed reasoning for their results
• Source links and references for research-based findings
• Confidence indicators and uncertainty acknowledgment
• Manual override capabilities when needed

​

Interactive Chat

• Understand your datasets, tools, and ongoing projects
• Access to workspace context for informed conversations
• Ability to help create new columns, records, and structures
• Integration with your existing data and workflows

• Use connected tools during chat conversations
• Research information, verify facts, and gather current data
• Transparent tool usage with expandable details
• Real-time tool execution with status updates

• Copy and share conversation results
• Build on previous interactions and context
• Help with data analysis, planning, and decision-making
• Support for both quick questions and deep exploration

​

Creating and Configuring Agents

​

Basic Information

• Name: Descriptive names like “Company Research Agent” or “Document Analyzer”
• Description: Detailed explanation of the agent’s purpose and capabilities
• Status: Active or inactive state for workspace management

​

Model Configuration

• OpenAI: GPT-4o, GPT-4o Mini, GPT-4.1 series, GPT-5 series
• Anthropic: Claude Opus 4.1, Claude Sonnet 4.0, Claude Haiku variants
• Future Models: System designed to support new model releases

• Use workspace variables for secure credential storage
• Never paste API keys directly into agent configuration
• Automatic validation of API key functionality
• Support for multiple API keys across different providers

​

System Prompt Engineering

You are a professional business research analyst specializing in
company intelligence and market research.

When provided with company information, your task is to:
1. Research and analyze the company's core business
2. Determine the primary industry and market sector
3. Estimate company size and maturity level
4. Identify key products or services offered
5. Assess market position and competitive landscape

Provide your analysis in JSON format:
{
  "industry": "Primary industry classification",
  "company_size": "Estimated size (Startup, Small, Medium, Large)",
  "primary_products": ["List of main products/services"],
  "confidence_score": 0.85
}

​

Tool Selection

• Select relevant tools from your workspace tool library
• Tools are filtered and searchable by name, type, and description
• Monitor tool usage and performance in chat conversations

​

Agent Management

​

Agent Dashboard

• Agent List: View all agents with status, model, and activity information
• Search and Filter: Find agents by name, model, or creation date
• Quick Actions: Edit, duplicate, test, and chat with agents
• Performance Metrics: Success rates, usage statistics, and processing history

​

Agent Operations

• Step-by-step creation wizard with validation
• Real-time preview of configuration
• Tool selection with capability previews
• API key validation and testing

• Interactive chat interface for testing
• Tool usage verification in controlled environment
• Prompt refinement through iterative testing
• Performance evaluation with sample data

• Update system prompts and tool selections
• Modify model configurations and API keys
• Manage agent status and availability
• Version control for configuration changes

​

Team Collaboration

• Agents are available to all workspace members
• Consistent processing across team workflows
• Shared context and institutional knowledge
• Collaborative prompt engineering and refinement

• Workspace-level agent management
• Tool permissions and usage controls
• API key security and access management
• Usage monitoring and cost tracking

​

Next Steps