Argilla
argilla.io
- Free plan
- Linux
- Web
- API
- Self-hosted
What Argilla is
Argilla is a free, open-source collaboration tool for AI engineers and domain experts building datasets with human feedback. It supports NLP, LLM, and multimodal projects, including text classification, named entity recognition, retrieval-augmented generation, and preference tuning. Contributors can submit labels, ratings, rankings, or text responses. Annotation tools include filters, AI feedback suggestions, and semantic search. Argilla handles text and images, while custom fields can represent audio, video, or other material rendered as base64 or HTML. Datasets and records can be imported from or exported to Python, local disk, or the Hugging Face Hub. It is compatible with Hugging Face and spaCy, with tutorials for SetFit and LlamaIndex. Teams can deploy it on Hugging Face Spaces or on a local machine or server using Docker Compose. A Python SDK connects to an Argilla server through an API URL and key, and the server exposes REST API documentation. OAuth2 sign-in is available through Hugging Face, GitHub, and Google. Argilla does not train models; its FAQ points users to a separate machine-learning framework. On Hugging Face Spaces, data on ephemeral free storage is lost when the Space restarts.
Who it suits
Argilla suits AI teams that need domain experts to contribute and review dataset feedback. It supports both hosted Spaces and local or server deployment.
Good to know, and what to watch
- Collects labels, ratings, rankings, and written responses.
- Annotation tools include filters and semantic search.
- Datasets can move between Python, local disk, and Hugging Face Hub.
- Offers a Python SDK and REST API documentation.
- Supports OAuth2 through Hugging Face, GitHub, and Google.
- Argilla does not train models.
- Ephemeral free storage on Hugging Face Spaces is lost on restart.
- Support is directed to the community.
Plans
Every plan and priceArgilla
Free plan
Getting started
- Open the Argilla website
- Deploy on Hugging Face Spaces or a local machine or server with Docker Compose
- Configure OAuth2 with Hugging Face, GitHub, or Google if desired
- Connect the Python SDK using an Argilla server API URL and API key
- Import or create datasets through Python, local disk, or the Hugging Face Hub
Limits to check
Argilla is free and open source, but data on ephemeral free storage in Hugging Face Spaces is lost when the Space restarts. The software provides dataset workflows and does not train models.
The verdict
Argilla focuses on collaborative dataset creation rather than model training. Choose a separate machine-learning framework for training, and account for the Spaces storage limit if deploying there.
Questions people ask
Is Argilla free?
Yes. The Argilla plan costs nothing and is open-source software.
Where can Argilla run?
It can run on Hugging Face Spaces or on your own machine or server using Docker Compose. Listed platforms include API, Linux, self-hosted, and web.
Does Argilla train machine-learning models?
No. Its FAQ recommends using a separate machine-learning framework such as Hugging Face Transformers.
Which data types can it annotate?
Argilla supports text and images. Custom fields can represent audio, video, or other data rendered as base64 or HTML.
How does authentication work?
OAuth2 authentication is supported with Hugging Face, GitHub, and Google providers by default.
Can datasets move between systems?
Yes. Datasets and records can be imported from and exported to Python, local disk, or the Hugging Face Hub.
What the maker publishes
- AI workflows
- It supports collecting human feedback for NLP, LLM, and multimodal projects, including tasks such as text classification, named entity recognition, retrieval-augmented generation, and preference tuning. source
- Annotation tools
- Users can label data with filters, AI feedback suggestions, and semantic search. source
- Authentication
- Argilla supports OAuth2 authentication with Hugging Face, GitHub, and Google providers by default. source
- Data formats
- Argilla supports text and images, and custom fields can represent audio, video, or other data rendered as base64 or HTML. source
- Data persistence limit
- On Hugging Face Spaces, data on ephemeral free storage is lost when the Space restarts. source
- Data portability
- Datasets and records can be imported from and exported to Python, local disk, or the Hugging Face Hub. source
- Deployment
- Argilla can be deployed on Hugging Face Spaces or on a local machine or server using Docker Compose. source
- Feedback types
- Datasets can collect feedback such as labels, ratings, rankings, and text responses. source
- Integrations
- The docs describe compatibility with Hugging Face and spaCy, and tutorials show integrations with SetFit and LlamaIndex. source
- Model training limit
- Argilla does not train models; its FAQ recommends using a separate machine-learning framework such as Hugging Face Transformers. source
- Privacy and telemetry
- Argilla reports anonymous usage and error telemetry, says it does not collect dataset records, names, or metadata, and allows telemetry to be disabled with an environment variable. source
- Purpose
- Argilla is a collaboration tool for AI engineers and domain experts to build high-quality datasets. source
- SDK and API
- A Python SDK connects to an Argilla server using its API URL and API key, and the server exposes REST API documentation. source
- Support
- The product site directs users to its community for support and use-case discussion. source
Checked against the maker's pages on 30 Sept 2026.