doccano

Overview

doccano is open-source software for labeling data used in machine-learning work. It suits teams that want to organize annotation projects themselves and export the results into their own workflows.

Its appeal is a free, adaptable workflow with collaboration, model assistance, and REST APIs. The trade-off is operational responsibility: teams choose and maintain their own deployment and database, and SQLite upgrades carry a risk of database loss.

Key features

Projects support text classification, sequence labeling, and sequence-to-sequence tasks. Examples include sentiment analysis, named-entity recognition, and text summarization, making doccano a strong fit for teams structuring text labels rather than seeking a broad project-management suite.

A project can take a team from dataset import through user setup, annotation guidelines, labeling, and export. Collaborative annotation and human review workflows suit shared labeling efforts; custom ontologies and model-assisted labeling add flexibility when teams need tailored categories or machine-learning assistance. Supported modalities include text, image, and audio, although the documented task types focus on text.

REST APIs connect doccano with scripts and machine-learning models, which is useful when labeling needs to feed an existing pipeline. Multi-language and mobile support, emoji support, and a dark theme round out the workflow, but do not change its central role as an annotation tool.

Pricing

doccano is free, open-source software under the MIT license. The MIT-licensed software plan costs 0.00 USD per free, with permission to use, copy, modify, publish, distribute, sublicense, and sell the software. There is no paid tier identified; the practical cost is the effort of choosing, deploying, and maintaining the installation.

Platforms

doccano is available for Android, iOS, Linux, macOS, Windows, web, and self-hosted use, with API access. Installation is documented through pip, Docker, and Docker Compose. The project documents installation on Linux, Windows, or macOS machines running Python 3.8 or newer, plus one-click deployment options for AWS and Heroku and Docker deployment elsewhere.

Imported datasets can use Amazon S3 or Google Cloud Storage. SQLite 3 is the default database, while PostgreSQL and other database systems are also described. Celery handles long-running imports and exports; documented message-broker options include SQLite3, RabbitMQ, and Redis. The SQLite upgrade warning matters: an upgrade can lose the database, so teams should weigh that risk when choosing the default.

The repository has no detected security policy and no published security advisories. Documentation points users to the FAQ and says they can send help requests and feedback to the author. Teams that need a formal security policy or a defined support channel may prefer another option.

Who it's for

doccano is a good match for machine-learning practitioners and teams that want collaborative annotation, human review, custom ontologies, and API access without a software license fee. Its multiple installation routes and storage and database options suit teams prepared to manage their own deployment.

It is less suitable for teams that want vendor-managed operations or do not want to handle database and upgrade decisions. In particular, the SQLite upgrade caution is important for anyone considering the default database.

Pros and cons

• Pro: The free MIT-licensed software permits modification and redistribution, giving teams room to adapt their deployment and workflows.
• Pro: Collaboration, guidelines, review workflows, model assistance, and exports connect team labeling to downstream machine-learning work.
• Pro: REST APIs and installation choices including pip, Docker, and Docker Compose make it adaptable to different technical setups.
• Con: Teams take on deployment and database choices themselves, rather than relying on a managed service.
• Con: Upgrading a SQLite installation can lose its database, a material risk for teams using the default.
• Con: Its documented task types are text-focused even though the supported modalities also include image and audio.

Alternatives

AI Data Labeling Tools is a useful starting point for comparing options across the category.

Label Studio is worth considering for readers who want a freemium alternative with a free plan and trial, and support across web, API, desktop operating systems, and self-hosted deployment.

Alibaba Cloud PAI iTAG may suit teams looking for manual labeling at no platform charge; OSS storage and data transfer are billed separately.

BRAT is another free, MIT-licensed, self-hosted option for readers who want a Linux, macOS, or web-based tool.

Roboflow is a freemium alternative with a free tier of 10 credits a month, enough to train about 30 models or run 80,000 inferences. Its Core plan is 39.00 USD per month.

Amazon SageMaker Autopilot is an alternative for readers seeking a paid, pay-as-you-go service with a free trial; its listed on-demand pricing has no minimum fees or upfront commitments.

Argilla is another free, open-source choice, deployable on Hugging Face Spaces or a team’s own infrastructure.

CVAT is a freemium alternative with a free Community plan for personal use and small teams, as well as a free online plan limited to one member, one project, three tasks, and 1 GB.

Supervisely offers a free Community plan with 5 GB storage, 10,000 files, and two members, alongside a Pro plan at 199.00 EUR per month.

Verdict

Choose doccano if your team wants free, open-source annotation with collaborative workflows, model assistance, and APIs, and is prepared to operate the deployment. Look elsewhere if you need managed operations or want to avoid the SQLite upgrade risk.