No Coding Required

Coze's visual workflow editor lets you build AI agents by dragging and dropping components. No programming knowledge needed!

Rapid Development

Reduce development time by 90% compared to traditional coding. Turn your ideas into working AI agents in minutes.

Modular Components

Access a rich library of pre-built nodes including plugins, LLM integration, knowledge bases, and conditional logic.

Easy Deployment

Publish your AI agents to multiple channels including Coze platform, API, WeChat, and Feishu with just a few clicks.

Coze Platform Interface

The intuitive interface makes it easy to navigate and build your AI agents. Here's a preview of what you'll see:

Workflow

A predefined set of steps to complete a specific task. Like the famous "putting an elephant in a fridge" example - each step is a node connected in sequence.

Nodes

The basic building blocks of workflows. Each node performs a specific function. Coze provides 13 types including plugins, LLM, code, knowledge base, and conditional nodes.

Parameter Passing

The connection between nodes. There are two types: Reference parameters (reusing data) and Input parameters (user-provided data).

Understanding Flowcharts

Workflows are essentially visual flowcharts. Understanding how to read and create them is key to building effective AI agents:

Step 1: Register for a Coze Account

1. 1

   Visit the Coze official website

2. 2

   Click on the "Register" button in the top right corner

3. 3

   Register using your email address (recommended)

4. 4

   Complete the verification process (takes about 2 minutes)

Step 2: Familiarize Yourself with the Interface

After logging in, you'll see the main workspace, which is divided into four main areas:

Step 1: Create the Agent Framework

1. 1

   Click on "Agents" in the left navigation bar

2. 2

   Click the "+ Create Agent" button

3. 3

   Fill in the basic information:

• Name: WeatherBot (only letters, numbers, and underscores)

• Description: A weather assistant that checks real-time weather

• Icon: Generate with AI or upload your own

4. 4

   Click "Confirm" to enter the agent editing page

Step 2: Add Workflow and Configure Start Node

1. 1

   In the agent editing page, click the Workflow" tab

2. 2

   Click the "+" button to create a new workflow

3. 3

   Name it "weather_query" with description "Query weather for a specified city"

4. 4

   Configure the start node parameters in the right property panel:

• Parameter name: city_name (follow naming conventions)

• Type: Select "String"

• Description: Enter the city name to query

• Required: Check Yes"

Step 3: Add Function Nodes and Connect Them

Now we'll implement the core functionality: User inputs city name → Call weather API → Return weather information

1. 1

   From the left node library, find the "Plugin" node and drag it to the canvas

2. 2

   In the plugin list, select the Weather Query" plugin (install it first if not available)

3. 3

   Connect the nodes: Drag a line from the right dot of the start node to the left dot of the weather plugin node

4. 4

   Configure the weather plugin node:

• City parameter: Select "Reference Parameter" and choose city_name from the start node

• Temperature unit: Select "Celsius"

• Language: Select "English"

Step 4: Process Results and Optimize Output

The raw data returned by the weather plugin needs to be formatted into user-friendly language using an LLM node.

1. 1

   Add an LLM node and connect it to the weather node

2. 2

   Configure the prompt in the LLM node:

   You are a weather forecast assistant. Please convert the following weather data into a friendly Chinese response: City: ${weather_data.city} Temperature: ${weather_data.temperature}°C Weather condition: ${weather_data.condition} Humidity: ${weather_data.humidity}% Please respond in a warm and friendly tone and add appropriate clothing suggestions.

    

3. 3

   Note the variable reference syntax ${}, which allows you to use output from previous nodes as input for the current node

Step 5: Test and Debug

1. 1

   Click the "Test Run" button in the top right corner

2. 2

   Enter a test city (e.g., "Beijing") and check the results

3. 3

   Common troubleshooting:

• "Parameter missing" error: Check node connections and parameter references

• Empty weather data: Try using English or Pinyin for city names

• Poor LLM output: Adjust the tone and format requirements in the prompt

4. 4

   After successful testing, connect the LLM node output to the end node

Step 6: Publish and Deploy

1. 1

   Click the "Publish" button in the top right corner

2. 2

   Enter version notes and click "Publish"

3. 3

   Return to the agent editing page and select deployment channels in the "Publish" tab:

• Coze Platform: Users can find your agent in the Coze store

• API Interface: For integration with other applications

• WeChat/Feishu: Direct deployment to common office platforms

4. 4

   For first-time发布, recommend choosing the Coze platform for easy sharing and testing

Add Error Handling

Provide friendly prompts when weather queries fail. Add conditional nodes to check if returned data is valid and redirect to fallback responses if not.

Improve User Experience

Add welcome messages and closing remarks. Configure the start node with a greeting like Welcome to Weather Query Assistant! Please enter the city name you want to check."

Extend Functionality

Add support for multiple cities, weather alerts, and detailed clothing recommendations. This can be achieved by adding additional nodes to the workflow.

Advanced Workflow Example

As you become more comfortable with Coze, you can create more complex workflows with multiple branches and integrations:

Comprehensive guides and references for all Coze features and functionalities.

Step-by-step video guides for building various types of AI agents.

Connect with other developers, share your creations, and get help with problems.

Explore and modify pre-built agent templates to accelerate your development.