Skip to main content

​
What Are AI Connections in AgentFlow?

Understand how AgentFlow connects to AI and distributes it across organizations. AgentFlow is a connection and distribution platform for AI. You bring your own AI endpoints (OpenAI, custom workflows, agent builders, etc.), and AgentFlow provides the interface, access control, and multi-organization distribution.

​
Understanding the Platform

​
The Simple Explanation

Think of AgentFlow as infrastructure for AI deployment:
  • You create AI agents, workflows, or endpoints elsewhere
  • You connect them to AgentFlow via API
  • AgentFlow provides the chat UI, user management, and distribution
  • You deploy to multiple organizations with data isolation

​
What AgentFlow Provides

What You BuildWhat AgentFlow Provides
AI endpoint/agentChat interface
Custom logicUser management
Agent behaviorAccess control
Prompts & trainingDistribution layer
API responsesOrganization isolation
Business logicAnalytics & monitoring

​
Connection Components

Every AI connection in AgentFlow consists of:

​
1. The Endpoint URL

The AI service or workflow you’re connecting to:

AI Models

Direct API connections
  • OpenAI (GPT-4, GPT-3.5)
  • Anthropic (Claude)
  • Google (Gemini)
  • Custom AI services

Workflows

Automation platforms
  • N8N workflows
  • Make.com scenarios
  • Zapier automations
  • Custom logic flows

Agent Builders

Pre-built agent systems
  • ChatGPT Agent Builder
  • LangChain applications
  • Custom agent endpoints
  • OpenAI Assistants

Custom Endpoints

Any webhook service
  • Serverless functions
  • Cloud Run services
  • Microservices
  • HTTP APIs

​
2. Request/Response Mapping

How AgentFlow communicates with your AI connection: Request Format:
  • AgentFlow sends user messages to your endpoint
  • Uses configurable field mapping
  • Supports custom headers and authentication
  • Works with any JSON structure
Response Format:
  • Your endpoint returns the AI’s response
  • AgentFlow extracts the message content
  • Displays to the user in the chat interface
  • Maintains conversation history

​
3. Access Control

Who can use this AI connection:
  • Groups: Organize users by team or purpose
  • Permissions: Control which connections each group can access
  • Organization Isolation: Each org’s data stays separate
  • Distribution: Deploy connections across multiple organizations

​
Types of AI Connections

AgentFlow can connect to any AI endpoint. Here are common types:

​
Customer Support Agents

Connect to AI-powered support systems that help customers with questions and troubleshooting. Typical Connections:
  • OpenAI API with support knowledge
  • Custom workflow with CRM integration
  • Pre-built support agent APIs

​
Coding Assistants

Connect to technical AI systems that help with code, debugging, and technical documentation. Typical Connections:
  • GPT-4 API for code generation
  • LangChain agent with code tools
  • Custom coding endpoint

​
Content Writers

Connect to AI systems designed for creating marketing content, blog posts, and copy. Typical Connections:
  • Claude API for long-form content
  • Custom workflow with brand guidelines
  • Content generation endpoints

​
Data Analysts

Connect to AI systems that analyze data and generate reports. Typical Connections:
  • Custom endpoint with database access
  • Workflow automation + AI
  • Analytics-focused AI service

​
Multi-Step Workflows

Connect to complex automation systems that combine multiple services. Typical Connections:
  • N8N workflows with AI integration
  • Make.com scenarios
  • Custom business logic endpoints

​
What AgentFlow Provides

AgentFlow acts as the connection and distribution layer:
  • Clean chat UI for users
  • Conversation history
  • Multi-user support
  • Organization management
  • Connect any AI endpoint
  • Configure request/response mapping
  • Secure authentication handling
  • Test before deployment
  • Group-based permissions
  • User management
  • Role assignments
  • Audit trails
  • Deploy to multiple organizations
  • Data isolation per org
  • Centralized management
  • Scalable architecture
  • Usage tracking
  • Cost monitoring
  • Performance metrics
  • Export capabilities

​
Important Notes

AgentFlow is a connection platform. You connect your own AI agents, workflows, or endpoints to AgentFlow. The platform handles:
  • User interface and conversations
  • Access control and permissions
  • Distribution across organizations
  • Analytics and monitoring
What AgentFlow Doesn’t Do:
  • Build or train AI models
  • Create agent logic or prompts
  • Host AI inference
  • Provide AI capabilities directly
You bring your own AI endpoints, and AgentFlow distributes them.

​
Agent Distribution

One of AgentFlow’s most powerful features:

​
What is Distribution?

Deploy your agent to multiple organizations while maintaining complete data isolation. 
     Your Agent
         ↓
    ┌────┴────┐
    ↓         ↓
  Org A     Org B
    ↓         ↓
  Data A    Data B
  (isolated)(isolated)

​
Why Distribute?

SaaS Companies:
  • Provide AI to every customer
  • Each customer gets isolated instance
  • Customize per customer
Agencies:
  • Deploy same agent to all clients
  • Maintain separate data
  • Centralized improvements
Enterprises:
  • Distribute across departments
  • Maintain security boundaries
  • Consistent capabilities

Learn About Distribution

Complete guide to agent distribution

​
Best Practices

​
1. Start Simple

Don’t over-configure your first connection:
  • Start with a simple AI API
  • Test with one endpoint
  • Add complexity as needed
  • Verify it works before distributing

​
2. Test Thoroughly

Before deploying to users:
  • Test your endpoint independently
  • Verify request/response mapping
  • Try various questions and scenarios
  • Get feedback from team members

​
3. Monitor Performance

Track how your connections perform:
  • Review conversation logs
  • Monitor response times
  • Track costs and usage
  • Gather user feedback

​
4. Iterate and Improve

Connections improve over time:
  • Update endpoints as needed
  • Optimize response mapping
  • Refine access controls
  • Scale based on usage patterns

​
Real-World Examples

​
Example 1: SaaS Company Support

Use Case: Deploy AI support agent to 500+ customers Setup:
  • Connected GPT-4 API endpoint
  • Distributed to all customer orgs
  • Each org’s data isolated
  • Centralized updates
Benefits:
  • One configuration, 500 deployments
  • Update once, affects all customers
  • Complete data isolation
  • Easy to manage

​
Example 2: E-commerce with Inventory

Use Case: Product recommendations with real-time inventory Setup:
  • Custom N8N workflow endpoint
  • Checks inventory database
  • Calls AI for recommendations
  • Returns personalized results
Benefits:
  • Complex logic handled in workflow
  • Real-time data integration
  • Custom business rules
  • Flexible and powerful

​
Example 3: Multi-Department Enterprise

Use Case: Different AI tools per department Setup:
  • HR: Claude API for onboarding
  • Sales: Custom CRM + AI workflow
  • Support: GPT-4 API
  • Each in separate groups
Benefits:
  • Appropriate AI per department
  • Granular access control
  • Separate cost tracking
  • Unified interface

​
What’s Next?

Ready to create your own agent?

Create Your First Agent

Step-by-step agent creation guide

Choose an AI Connection

Compare AI connections and pick the right one

Configure Your Agent

Advanced configuration options

Test Your Agent

Best practices for testing
⌘I