Step-by-Step Local Deployment Guide for Coze Studio: Launch Your Low-Code AI Agent Development

Coze Studio is an open‑source low‑code AI Agent development platform built with a Golang backend and a React + TypeScript frontend, using a microservice and DDD architecture for high performance and extensibility.

1. Install Ollama (private LLM runtime)

Ollama is a lightweight framework for running large language models locally. Download the client from https://ollama.com/ for macOS, Linux or Windows and follow the installer.

Pull a model (e.g., Qwen 3) with one of the following commands:

ollama run qwen3:8b
# or
ollama run qwen3:14b
# or
ollama run qwen3:32b

Memory requirements:

7B ≈ 8‑16 GB RAM

14B ≈ 16‑32 GB RAM

32B ≈ 32‑64 GB RAM

70B+ ≈ 64 GB+ RAM

2. Install Docker

Docker provides containerised deployment. Download Docker Desktop from https://www.docker.com/ and install the graphical client for your OS.

3. Deploy Coze Studio locally

3.1 Environment requirements

Minimum 2 CPU cores, 4 GB RAM.

Docker and Docker Compose installed; Docker daemon running.

3.2 Get the source code

git clone https://github.com/coze-dev/coze-studio.git

If Git is unavailable, download the zip archive from the project page.

3.3 Configure model templates

Copy a template from backend/conf/model/template to backend/conf/model. Example: rename model_template_ollama.yaml to model_ollama_qwen3_8b.yaml and set model: qwen3:8b. Ensure the id field is a non‑zero unique integer.

For an online service, copy model_template_deepseek.yaml to model_openrouter_ds_v3.yaml and set the DeepSeek‑V3 API key obtained from OpenRouter (

https://openrouter.ai/deepseek/deepseek-chat-v3-0324:free/api

).

Use ollama list to verify locally installed models.

3.4 Start the services

cd docker
cp .env.example .env
docker compose --profile '*' up -d

The first run pulls and builds images; wait until the coze-server container turns green.

To apply configuration changes, restart the server:

docker compose --profile '*' restart coze-server

3.5 Use Coze Studio

Open a browser at http://localhost:8888/, register with an email and password, then create an agent via the “Create” button. The model list shows the configured LLMs; test them in the chat pane.

If errors occur, inspect the coze-server logs.

Project repository:

https://github.com/coze-dev/coze-studio