Coze Studio - ByteDance's Open Source AI Agent Development Platform

What is Coze Studio

Coze Studio (Coze Development Platform) is an open source project by ByteDanceAI agent development platform, helping developers quickly build, deploy, and manage AI agents. It provides a one-stop development environment, supporting core technologies such as Prompt, RAG, Plugin, and Workflow. Through visual workflow orchestration, developers can build complex AI applications with zero code or low code. The platform has a built-in plugin framework that can encapsulate third-party APIs or private capabilities as plugins to extend agent functionality. Coze Studio is developed with Golang and React, follows Domain-Driven Design (DDD) principles, and is easy to customize. It supports local and private deployment with a low deployment threshold, requiring only a dual-core CPU and 4GB of memory to run, and provides a one-click deployment script.

Main Features of Coze Studio

• One-stop development environment: Provides full-process support from development to deployment, covering core technologies such as Prompt, RAG, Plugin, and Workflow.
• Visual workflow orchestration: Quickly build complex workflows by dragging and dropping nodes, supporting zero-code or low-code development, making it convenient for developers to design business logic.
• Agent building and management: Provides robust application templates and orchestration frameworks, supporting rapid creation, publishing, and management of AI agents.
• Multi-model support: Supports integration of multiple mainstream large models (such as OpenAI, Volcano Engine, etc.) and provides a unified model interface abstraction.
• Knowledge base management: Integrates RAG (Retrieval-Augmented Generation) capabilities, supports vector retrieval, allows uploading documents or data, and enables agents to generate accurate answers based on the knowledge base.
• Plugin system: Supports creating, configuring, and managing plugins, can encapsulate third-party APIs or private functions, and extends agent capabilities.
• Database support: Integrates database resources and supports data storage and querying.
• API and SDK integration: Provides chat and workflow-related APIs, supports SDKs for Python, JavaScript, Java, etc., facilitating secondary development.
• Real-time interaction: Supports real-time chat, speech synthesis, and transcription via WebSocket.

Coze Studio Project Address

• Github repository：https://github.com/coze-dev/coze-studio

How to Use Coze Studio

• Environment preparation
  • Ensure the computer meets the minimum configuration requirements: dual-core CPU and 4GB memory.
  • Install Docker and Docker Compose in advance, and start the Docker service.
• Get the source code: Open a terminal or command prompt and run the following command to clone the Coze Studio source code: git clone https://github.com/coze-dev/coze-studio.git
• Configure the model
  • Enter the cloned project directory: cd coze-studio
  • Copy the model configuration template file from the template directory: cp backend/conf/model/template/model_template_ark_doubao-seed-1.6.yaml backend/conf/model/ark_doubao-seed-1.6.yaml
• Modify the configuration file：
  • Enter the backend/conf/model directory.
  • Open the ark_doubao-seed-1.6.yaml file.
  • Set the following fields:
    • id: Model ID, defined by the developer, must be a non-zero integer and globally unique.
    • meta.conn_config.api_key: API Key for the model service.
    • meta.conn_config.model: The model ID of the model service.
• Deploy and start the service
  • Enter the docker directory, copy the environment variable template file, and start the service.
  • The first deployment may take a long time because it may need to pull images and build local images. If you see the prompt "Container coze-server Started", the service has started successfully.
• Access the platform: Open a browser and visit http://localhost:8888/to open the Coze Studio frontend page.
• Register and log in: On the Coze Studio login page, register and log in with your email or phone number.
• Create an agent：After logging in, click the "Create Bot" button in the upper left corner and enter basic information about the Bot (such as name, avatar, etc.).
• Configure the agent：
  • On the configuration page, you can set the Bot's role, skills, and limitations (prompts).
  • Add official plugins or custom plugins, and configure workflows, knowledge bases, or databases.
  • Converse with the Bot on the debugging page to test its functionality.
• Use plugins
  • Use plugins in a Bot：
    • On the Bot editing page, select the plugin area and click the plus icon to add a plugin.
    • Configure plugin parameters and test the plugin functionality in the preview and debugging area.
  • Use plugin nodes in a workflow：
    • Create or select a workflow, and add a plugin node to the workflow.
    • Configure input and output parameters and test the plugin functionality.
• Publish the application: After completing all configurations, click the "Publish" button in the upper right corner to publish the agent to the Coze app store.

Coze Studio application scenarios

• Rapid prototypingvalidation (POC): Developers can quickly build prototypes of AI applications and validate ideas and concepts.
• AI application implementation for small and medium-sized enterprises: Small and medium-sized enterprises can use Coze Studio to quickly leverage large model capabilities to solve business problems.
• Private deployment: For enterprises with high data security requirements, Coze Studio supports private deployment to ensure that business data does not leave the enterprise intranet.
• Enterprise internal Q&A robot: Use a knowledge base to build a company SOP agent, and add approval, form filling, email sending, and other logic through workflows.
• AI interactive questionnaire system: Use workflows to guide users in filling out questionnaires, call plugins to generate data charts, and synchronize the data to the backend BI system.