---
id: cvat-sdk
version: "2.73.0"
license: MIT
license_treatment: permissive
maintenance: active
---

# cvat-sdk — Software Development Kit for CVAT

License: permissive · Maintenance: active · Downloads: 104.6K/mo

## What it is and what it does

cvat-sdk is a Python client library that connects your code to a running CVAT server, allowing you to automate annotation workflows, retrieve and manage annotation data, and integrate computer vision annotation into larger pipelines. It exposes multiple API layers: low-level server API wrappers for direct control, high-level tools for common tasks, a PyTorch adapter for deep learning workflows, and auto-annotation support for model-driven labeling. The library depends on standard utilities like attrs, Pillow, tqdm, and urllib3, making it lightweight to install.

The package is actively maintained (latest release 2 days old) with strong community backing (16525 GitHub stars). It supports current Python versions (3.10+) and is distributed as a pure wheel, so installation is straightforward. Optional extras allow you to pull in PyTorch or mask-processing dependencies only when needed. No known security vulnerabilities are recorded.

Use it for:

- Automate bulk annotation uploads and downloads from CVAT without manual UI interaction.
- Build custom labeling workflows that integrate CVAT with your training pipeline.
- Use PyTorch models for auto-annotation to speed up large-scale annotation projects.
- Query and filter annotation metadata programmatically for quality assurance or analysis.
- Integrate CVAT into CI/CD pipelines for continuous dataset generation and validation.

## Worth the install?

AI-flagged interpretation of the facts above — verify before relying.

Python client library for CVAT (Computer Vision Annotation Tool) server, enabling programmatic access to annotation workflows, data management, and auto-annotation features through a layered API.

Yes. The package is actively maintained, has no known vulnerabilities, installs with low friction, and carries a permissive MIT license. It is the official SDK for CVAT and well-suited for teams automating annotation workflows or integrating CVAT into larger systems. Install it if you need programmatic access to a CVAT server; skip it if you only use CVAT's web UI.

## Install

pip install cvat-sdk
uv add cvat-sdk
poetry add cvat-sdk

## Installing cvat-sdk

Before you install: Low friction installation with a pure-wheel distribution. Active maintenance with a recent release (2 days old) and a well-established repository (16525 stars). Requires Python 3.10 or later.

License in practice: MIT license permits free use, modification, and distribution with minimal restrictions, making it suitable for both open-source and commercial projects.

Quickstart:

pip install cvat-sdk

import cvat_sdk
# Access CVAT server through the high-level Core API
from cvat_sdk.core import Client

Requires Python 3.10 or later; a CVAT server must be running and accessible.

Verify before relying:

- Whether the PyTorch and masks extras are commonly needed or optional for typical workflows.
- Performance characteristics when handling large annotation projects or high-volume API calls.
- Compatibility guarantees with specific CVAT server versions.
- Specific server address and authentication requirements for Client initialization.

## Package facts

- License: MIT (permissive)
- Python support: supports_current
- Install friction: low
- Maintenance: active
- Downloads: 104.6K/month (top 15,000 on PyPI)
- Known vulnerabilities: none known

## Tags

CVAT Python client, computer vision annotation API, annotation workflow automation, CVAT server integration, image annotation SDK, PyTorch annotation adapter, auto-annotation framework, computer-vision, annotation-tool, workflow-automation

[View on SkillFed](https://skillfed.io/packages/cvat-sdk) · [View on PyPI](https://pypi.org/project/cvat-sdk/)