Skip to main content
Chatflows are visual workflows that connect AI models, tools, and logic to create intelligent applications. This guide walks you through creating your first chatflow.

​
Understanding Chatflows

A chatflow is a visual representation of how data flows through different nodes:
  • Nodes: Individual components (LLMs, tools, memory, etc.)
  • Edges: Connections between nodes that define data flow
  • Canvas: The visual workspace where you build your flow

​
Creating a New Chatflow

1

Navigate to Chatflows

Click on Chatflows in the main navigation to view all your chatflows.
2

Start a New Flow

Click the Add New button to create a blank chatflow, or choose from templates in the marketplace.
3

Name Your Chatflow

Click the chatflow name in the header to edit it. Choose a descriptive name like “Customer Support Bot” or “Document QA”.

​
Building Your First Flow

Let’s create a simple conversational chatbot:
1

Add a Language Model

  1. Click the + Add Node button
  2. Navigate to Chat Models
  3. Select ChatOpenAI
  4. Configure your API credentials
  5. Set model parameters (temperature, max tokens, etc.)
2

Add Memory

Memory enables your chatbot to remember conversation history:
  1. Click + Add Node
  2. Go to Memory category
  3. Select Buffer Memory
  4. Configure session settings (optional)
3

Add the Chain

The chain connects everything together:
  1. Click + Add Node
  2. Go to Chains category
  3. Select Conversation Chain
  4. This node will have input anchors for the LLM and memory
4

Connect the Nodes

Draw connections between nodes:
  1. Click and drag from ChatOpenAI output to Conversation Chain model input
  2. Connect Buffer Memory output to Conversation Chain memory input
  3. Connections appear as lines between nodes
Output anchors are on the right side of nodes, input anchors on the left. Flowise validates connections to ensure compatible types.

​
Working with Templates

Flowise provides pre-built templates for common use cases:
  1. Go to Marketplace in the navigation
  2. Browse available chatflow templates
  3. Click Use Template on any flow
  4. The template opens in a new canvas
  5. Customize nodes and connections as needed
  6. Save your customized version

​
Canvas Controls

The canvas provides several controls for navigation and organization:
ControlDescription
Zoom In/OutUse mouse wheel or zoom controls
PanClick and drag on empty canvas area
Fit ViewAuto-fit all nodes into view
Snap to GridToggle grid snapping with magnet icon
BackgroundToggle grid background visibility
Mini MapOverview of entire flow (on large flows)

​
Configuring Chatflow Settings

Access via the settings icon in the header:
  • Category: Organize chatflows by category
  • Tags: Add searchable tags
  • Session Settings: Configure session timeout and behavior
  • Rate Limiting: Set request limits per user/session
  • Speech to Text: Enable voice input
  • Follow Up Prompts: Auto-generate suggested questions
  • Security: Configure allowed domains and authentication

​
Saving and Deploying

1

Save Your Flow

Click the Save button in the header. The flow is auto-saved to the database.
Unsaved changes are marked with a dirty indicator. You’ll be prompted before navigating away.
2

Test Your Flow

Click the Chat button to open the test interface. Send messages to verify behavior before deploying.
3

Deploy

Toggle the Deploy switch in the chatflows list. Deployed flows are accessible via API and embedded chat.

​
Common Patterns

​
Basic LLM Chain

Use for: Stateless prompt completion, text generation, classification

​
Conversational Agent

Use for: Chatbots with tool access, multi-step reasoning, dynamic actions

​
RAG (Retrieval Augmented Generation)

Use for: Document Q&A, knowledge bases, semantic search

​
Best Practices

  • Use Sticky Notes to add comments and documentation
  • Group related nodes together visually
  • Use consistent naming conventions
  • Keep flows under 15-20 nodes when possible
  • Break complex logic into multiple chatflows
  • Use appropriate model sizes (don’t default to largest)
  • Configure reasonable token limits
  • Implement caching for embeddings and LLM responses
  • Use streaming for better perceived performance
  • Store API keys in credentials, not hardcoded
  • Configure input moderation for user-facing bots
  • Set rate limits to prevent abuse
  • Use allowed domains to restrict access
  • Review system prompts for prompt injection risks

​
Next Steps

Working with Nodes

Learn about different node types and configurations

Testing & Debugging

Debug and optimize your chatflows

Variables & Expressions

Use dynamic values and expressions

Deployment

Deploy your chatflow to production

Build docs developers (and LLMs) love

Get started for freeTalk to us