Potato

Overview

Potato is open-source software for annotating text, images, audio, video, and qualitative data. It suits teams that need varied task types and can operate their own software. Its wide coverage and included AI assistance stand out in a free package, but self-hosting is mandatory.

Key features

Tasks are configured in YAML, and the maker says an annotation interface can be built without programming. That can help teams shape task-specific workflows without writing interface code; it does not eliminate the need to install and maintain Potato. Its 61 annotation types include classification, text spans, bounding boxes, polygons, video, audio, and qualitative coding, making it a strong fit for projects that span several data formats.

Potato also supports annotating multi-agent teams on an interaction graph and evaluating computer-use, voice, and video agents. This extends its usefulness beyond ordinary dataset labeling to teams examining agent behavior.

AI assistance includes label suggestions, keyword highlighting, pre-annotation, explanation generation, quality checking, and consistency checking. Integrations include OpenAI, Anthropic Claude, Google Gemini, Ollama, HuggingFace, OpenRouter, vLLM, YOLO, and LangChain. Together these give teams several routes to model-assisted labeling, while review workflows provide a way to check the resulting annotations.

For crowdsourced projects, Potato documents Prolific and Amazon Mechanical Turk integrations, with participant tracking and quality or approval workflows. Imports cover text, image, audio, video, and PDF or HTML documents; exports include JSON, JSONL, CSV, CoNLL, Hugging Face, COCO, and YOLO. That range is useful when annotation work needs to connect to different collection and downstream formats.

Potato runs locally or on users’ own servers, which the maker says keeps data on their infrastructure and supports offline use. It is licensed under GNU GPL version 3 or later, permitting use, modification, redistribution, and commercial use subject to the license terms. Installation starts with pip and requires Python 3.7 or later. Questions can go to Discord or the documentation, with a team contact email also provided.

Pricing

Potato is free: the Free plan costs nothing, is billed Free, and includes all features with no paid tiers or usage limits. That makes it a compelling option for teams that can handle deployment themselves and do not need a managed service. The absence of usage caps is generous, but it does not offset the operational burden for teams without the capacity to run their own instance.

Platforms

Potato supports API access and runs on Linux, macOS, and Windows, with web and self-hosted deployment options. Its quick start uses pip and requires Python 3.7 or later. The practical distinction is deployment: teams retain control of the environment, but must install and operate it themselves.

Who it's for

Potato is best for research and data teams handling multiple annotation formats, especially those that want AI-assisted labeling, crowdsourcing integrations, or workflows for evaluating agents. It is less suitable for teams seeking a hosted service or anyone unwilling to maintain self-hosted software.

Pros and cons

• Broad annotation coverage: 61 annotation types span text, images, audio, video, and qualitative coding, reducing the need to split varied work across separate tools.
• AI tools included at no cost: Suggestions, pre-annotation, and consistency checks can support labeling workflows, with review workflows available for checking annotations.
• Local and self-hosted operation: Keeping data on users’ infrastructure and supporting offline use suits privacy-conscious or disconnected environments.
• Self-hosting is required: Users must install and maintain Potato themselves; there is no managed cloud tier for teams that want someone else to run it.
• Installation has a technical prerequisite: The quick start requires pip and Python 3.7 or later, so teams need an environment capable of running it.

Alternatives

Data Labeling Software is a useful starting point for comparing tools across the category.

• Doccano is another free, open-source annotation tool installable with pip, Docker, or Docker Compose; consider it if you want a simpler deployment choice among those documented options.
• Label Studio has a Community Edition and a free trial; consider it if you want to compare a freemium alternative.
• LightlyStudio offers a free open-source version under Apache License 2.0; consider it if that license is a better fit.
• MakeSense.ai is a free, browser-based photo-labeling tool; choose it for photo labeling rather than Potato’s broader annotation scope.
• VGG Image Annotator is free, released under the two-clause BSD license, and has no paid tiers or usage limits stated; consider it if that licensing option suits your needs.
• Argilla is free open-source software deployable on Hugging Face Spaces or your own infrastructure; consider it if you want those deployment options.
• Roboflow has a free tier with 10 credits a month, enough to train about 30 models or run 80,000 inferences; choose it if that model-training or inference allowance matches your workload.
• Amazon SageMaker Autopilot is a paid, pay-as-you-go alternative with a free trial; consider it if that pricing approach suits your project.

Verdict

Choose Potato if you need broad annotation and AI-assisted labeling without usage limits, and can run the software on your own infrastructure. Its main advantage is the range of tasks and integrations in a free, self-hosted package; look elsewhere if you need managed hosting or do not want to take responsibility for deployment and maintenance.