Coze Studio is a unified development environment designed for building AI agents. It includes integrated support for modern large language models (LLMs), a comprehensive toolkit for various development patterns, and a streamlined path from initial prototype to production.
For developers creating complex AI agents, Coze Studio provides ready-made templates and a structured framework. This enables faster development cycles and lower overhead, allowing you to transform an idea into a functional agent without manually rebuilding underlying infrastructure for every project.
The platform simplifies the creation, debugging, and deployment of AI agents. Its visual design canvas supports low-code or no-code development, making it easy to assemble agents, applications, and workflows. You can also inject custom business logic where necessary. If your goal is to build low-code AI products, Coze Studio serves as an ideal starting point.
Coze Studio is powered by a Golang backend, while the frontend is built using React and TypeScript. The architecture relies on microservices and adheres to Domain-Driven Design (DDD) principles.
Feature Modules
| Module | Function |
|---|---|
| Model Service | Manages model inventories; integrates with OpenAI, Volcano Engine, and other providers. |
| Agent Builder | Builds, publishes, and manages agents; configures workflows, knowledge bases, and resources. |
| App Builder | Creates and publishes applications; defines business logic through custom workflows. |
| Workflow Builder | Provides tools to create, modify, publish, and delete automated workflows. |
| Resource Development | Manages plugins, knowledge bases, databases, and prompt engineering. |
| API & SDK | Offers OpenAPI endpoints for chat and conversation; allows embedding via the Chat SDK. |
Before installing Coze Studio, ensure your hardware meets the following minimum specifications: 2 CPU cores and 4 GB of RAM. Docker and Docker Compose must be installed and active on your system.
1. Clone the repository
git clone https://github.com/coze-dev/coze-studio.git
2. Configure the model
Copy the configuration template for the Doubao-seed-1.6 model from the template directory into your active configuration folder.
cd coze-studio
cp backend/conf/model/template/model_template_ark_doubao-seed-1.6.yaml backend/conf/model/ark_doubao-seed-1.6.yaml
Next, edit the new file. Navigate to backend/conf/model and open ark_doubao-seed-1.6.yaml. Configure the following fields:
id: A unique model identifier within Coze Studio. Assign a non-zero integer. Do not modify this ID after the model has been deployed.meta.conn_config.api_key: Your API key for the model service. If using Volcano Engine Ark, refer to their documentation to generate a key.meta.conn_config.model: The endpoint ID for the model service. For Volcano Engine’s Doubao-seed-1.6, use the specific Endpoint ID found in your access point settings.3. Deploy and start services
The initial launch will pull the necessary images and build local containers. This process takes several minutes, and logs will appear in your terminal. Wait until the message Container coze-server Started appears.
cd docker
cp .env.example .env
docker compose --profile "*" up -d
Following the startup, certain setup containers will show an exit 0 status. This is normal behavior for the following: coze-elasticsearch-setup, coze-minio-setup, coze-mysql-setup-init-sql, and coze-mysql-setup-schema.
Once all services are operational, access the interface at http://localhost:8888/ in your web browser.
The open-source version of Coze Studio utilizes Personal Access Tokens (PAT) for authentication. It supports both chat and workflow APIs. Please note that commercial-grade features are not included in this version.
For comprehensive usage instructions, please visit the official Coze documentation center. Note that certain capabilities—such as advanced tone customization—are reserved for the commercial edition. Refer to the feature comparison list for a full breakdown of differences.
Dianman VPN: Free Trial, Unlimited Data & Zero Throttling
Tiny Qwen: A Clean PyTorch Implementation of Qwen3 and Qwen2.5-VL
Kode CLI: A Multi-Model AI Terminal Assistant for Smarter Development
AoxVPN 8.8 Member Day Sale | No-Log VPN Featuring IEPL Private Lines
AIPy: Execute Python via Natural Language Directly in Your Terminal
Zettlr Setup and Developer Guide (macOS, Windows, Linux)
How to Install and Use Vosk Offline Speech Recognition
OCode: Native AI Coding Assistant for Your Terminal (Ollama)
ConEmu: A Highly Customizable Windows Terminal with Tabs and Split Panes
AingDesk: Run Local AI Models and Build a Private Knowledge Base
Gmail AutoAuth MCP Server: Control Gmail via Claude Desktop
AutoGenLib: Generate Python Code on the Fly With OpenAI API