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

Questions About Agents?

Check our FAQ or contact support@agentflow.live