Skip to the content
Know your tech before you buy it. Makers: get listed from $49/mo

DagsHub

dagshub.com

  • Free plan
  • Windows
  • macOS
  • Linux
  • Web
  • API
  • Self-hosted
8.3 Club score Visit dagshub.com

What DagsHub is

DagsHub helps developers and teams manage machine-learning projects from data collection and curation through annotation, experimentation, and model management. A project workspace can hold code, data, experiments, models, pipelines, and annotations together, and the platform is designed for text, images, audio, video, documents, medical imaging, and binary files. It hosts data but does not provide compute resources, so projects use local, cloud, or edge infrastructure for that work. Experiment tools support run comparison, metric and parameter visualization, real-time progress monitoring, and reproducibility. Data Engine provides curation, visualization, quality review, lineage, hard-example mining, labeling, auto-labeling, and data streaming. Integrations include GitHub, Google Colab, DVC, MLflow, Jenkins, Label Studio, Google Drive, Weights & Biases, and Hugging Face, among others. The Individual plan is free, with limits on private tracked experiments, collaborators, and storage. Paid pricing starts at $99/user/mo billed annually; a 14-day trial is listed. Enterprise supports on-premises, VPC, air-gapped, and OpenShift deployment.

Who it suits

DagsHub suits developers and teams managing machine-learning datasets, experiments, and models. Teams that require compute from the platform should note that DagsHub hosts data but relies on their own infrastructure for compute.

Good to know, and what to watch

  • Workspace brings code, data, models, and experiments together.
  • Supports a wide range of data types.
  • Data Engine includes labeling and data streaming.
  • Offers MLflow and Label Studio API support.
  • Enterprise deployment includes VPC and air-gapped options.
  • DagsHub does not provide compute resources.
  • Individual plan limits private experiments to 100.
  • Individual plan allows two private collaborators.
  • Individual plan includes 20GB storage.

Enterprise

—

Individual

Free

Individual

Free

Team

$99/mo

Free plan · 14-day trial

Getting started

  1. Open https://dagshub.com/
  2. Start with the free Individual plan or choose a paid plan
  3. Connect supported tools such as GitHub, DVC, MLflow, or Label Studio
  4. Use the web interface, API, or a self-hosted deployment
  5. Bring local, cloud, or edge infrastructure for computation

Limits to check

The free Individual plan allows 100 tracked experiments in private repositories, two private collaborators, and 20GB of DagsHub Storage. Team includes 1TB or two million files and costs $119 per user/month; Enterprise uses a custom quote.

The Club's review

DagsHub is a project hub for machine-learning data, code, experiments, annotations, and models, with support for everything from text and images to medical scans and binary files. It suits teams that need versioned, traceable workflows around their own infrastructure, while organizations seeking an all-in-one hosted compute environment should look elsewhere.

Overview

DagsHub covers the lifecycle from data collection and curation through annotation, experimentation, and model management. A project can bring together repositories, datasets, pipelines, experiment runs, models, and annotations, giving technical and labeling work a shared home. Dataset versioning, artifact tracking, a model registry, run comparison, and API/SDK access make it relevant beyond a simple file repository.

The central trade-off is hosting scope. DagsHub stores and organizes data but does not provide compute, so training and processing remain on local, cloud, or edge infrastructure. That is a sensible fit for teams with established compute, but a poor match for buyers expecting notebooks or managed training in the same service.

Key features

Multimodal data and Data Engine

DagsHub is built for text, images, audio, video, documents, medical imaging, and other binary data. Its Data Engine adds pandas-like curation, visualization, quality review, lineage, hard-example mining, labeling, auto-labeling, and data streaming. Together, these capabilities are aimed at teams that must repeatedly inspect and improve datasets rather than merely store them.

Experiments and reproducibility

Experiment tracking supports metric and parameter visualization, run comparison, real-time progress monitoring, and reproducibility. This is useful when teams need to explain why one model or dataset version won, and when multiple contributors must follow the same history.

Connections and APIs

Official integrations span GitHub, Google Colab, DVC, MLflow, Jenkins, Label Studio, external storage, Git servers, webhooks, Google Drive, Weights & Biases, Hugging Face, Apple MLX-LM, PyCaret, Giskard, and New Relic. Full MLflow and Label Studio API support helps teams bring existing tracking and annotation workflows into DagsHub instead of replacing them wholesale.

Enterprise deployment

Enterprise installations can run on-premises, in a VPC, in air-gapped environments, or on OpenShift. DagsHub states adherence to ISO 27001 and ISO 9001 and supports custom OIDC and LDAP authentication for SSO, addressing regulated or tightly controlled deployments.

Pricing

PlanPriceWhat you getBest for
IndividualFree (billed $0 per user/month)Unlimited public repositories; unlimited private repositories for non-commercial use; 100 private tracked experiments; two private collaborators; 20GB DagsHub Storage; community support.Individuals and non-commercial projects validating a workflow.
Team$119 a month (billed $119 per user/month)Unlimited private repositories; multimodal annotation and auto-labeling; own storage; team RBAC; 1TB or 2 million files; priority support.Commercial teams needing private collaboration and labeling at a defined capacity.
EnterpriseCustom quotePetabyte-scale data management; cluster deployment; VPC, air-gapped, or on-premises operation; SSO/LDAP/OIDC RBAC; enterprise SLA and support.Large or regulated organizations with deployment and governance requirements.

The service is freemium, with paid plans from $99/user/mo when billed annually and a 14-day trial. Individual is generous for public work but its private allowance is restrictive for a commercial team: only two private collaborators, 100 private tracked experiments, and 20GB of storage. Team removes the non-commercial private-repository restriction and adds operational controls, but its per-user billing makes larger groups materially more expensive. Enterprise is the route for scale and controlled deployment, with pricing negotiated.

Platforms

DagsHub is available through web, API, and self-hosted deployment. The API and SDK support suits automation, while self-hosting and enterprise deployment options serve organizations that cannot place projects in a standard hosted environment.

Who it's for

Choose DagsHub if your team needs one workspace linking datasets, annotations, code, experiments, pipelines, and models, especially across multimodal data and existing MLflow or Label Studio workflows. It is strongest for practitioners who already have compute and want governance, lineage, and collaboration around that compute. It is less suitable for a solo commercial user whose private project quickly exceeds the Individual limits, or for anyone wanting built-in training infrastructure.

Pros and cons

Pros

  • Broad ML lifecycle coverage: data curation, labeling, tracking, versioning, artifacts, and model registry share one project space.
  • Strong multimodal workflow: supports varied file types and adds Data Engine tools for review, lineage, and hard-example mining.
  • Integration depth: MLflow and Label Studio API support plus many official connectors reduce migration friction.
  • Deployment flexibility: self-hosted, VPC, air-gapped, and OpenShift options cover demanding environments.

Cons

  • No managed compute: users must supply local, cloud, or edge resources for processing and training.
  • Individual private caps: two collaborators, 100 private experiments, and 20GB storage constrain commercial use.
  • Team cost scales by user: $119 per user each month can become significant as a group grows.

Alternatives

Data Version Control Tools are the broader category to compare when dataset versioning is your primary requirement. For experiment-focused choices, see ML Experiment Tracking Software, while Model Registry Software is the relevant shortlist for model governance.

  • Weights & Biases is a strong alternative when you want a freemium tracking platform available across web, desktop, mobile, API, and self-hosted environments, with a Free plan offering five model seats, 5GB/month storage, and 1GB/month Weave ingestion.
  • Bauplan fits teams wanting a shared environment with CLI, SDK, and API access, provided public-by-default Sandbox projects are acceptable.
  • Graviti is worth considering when private cloud storage and row-based dataset capacity are the priority; its Starter plan includes 100GB and 50,000 rows.
  • Oxen suits users seeking unlimited public repositories and collaborators with a free Explorer plan that also permits five private repositories.
  • Apache Paimon is the free, open-source data-lake option for teams prepared to choose and operate their own connectors and storage plugins.
  • ClearML is preferable when an open-source self-hosted platform is central to the decision.
  • Databricks Notebooks is the better direction when managed notebook compute matters; its Free Edition includes one serverless workspace with limited compute size and usage.
  • DataLad is the free and open-source choice for users who need a platform available on Linux, macOS, and Windows.

Verdict

DagsHub is a strong choice for ML teams that need multimodal data management, annotation, reproducible experiment tracking, and model governance in one collaborative project space while retaining control of compute. Choose Team for commercial collaboration and Enterprise for governed, large-scale deployment. Look elsewhere if built-in compute is essential or if your private projects will outgrow the Individual plan's collaborator, experiment, or storage limits.

The verdict

DagsHub brings data management, annotation, experiment tracking, and model management into a shared project space. Its lack of built-in compute and the Individual plan's private-project limits are important considerations.

Questions people ask

Is there a free plan?

Yes. Individual costs $0 per user/month and includes unlimited public repositories, unlimited private repositories for non-commercial use, 100 private tracked experiments, two private collaborators, and 20GB of storage.

What does the paid Team plan cost?

Team is $119 per user/month, billed monthly. It adds unlimited private repositories, multimodal annotation and auto-labeling, own storage, team RBAC, and 1TB or two million files.

Is a trial available?

Yes. DagsHub lists a 14-day trial.

Which access options are available?

DagsHub is available through web, API, and self-hosted options. Enterprise installations can run on-premises, in a VPC, in air-gapped environments, or in OpenShift.

Does DagsHub provide compute?

No. It hosts data, while computation runs on the user's local, cloud, or edge infrastructure.

Does it support APIs and integrations?

It supports the MLflow and Label Studio APIs for hosted projects, with integrations including GitHub, Google Colab, DVC, Jenkins, Google Drive, Hugging Face, and others.

What the maker publishes

Annotation
The platform supports multimodal annotation and auto-labeling, with Label Studio compatibility on the Team plan. source
API
DagsHub publishes an API with interactive endpoint testing, token authentication and Basic Authentication using a username and token. source
API support
DagsHub offers full support for the MLflow and Label Studio APIs for projects hosted on DagsHub. source
CLI
The client supplies a dagshub command-line interface for login, repository creation, upload, download, DVC setup and version reporting. source
Client licence
The public DagsHub client repository is marked MIT licensed. source
Company founders
Crunchbase names Dean Pleban as co-founder and CEO and Guy Smoilovsky as CTO and co-founder. source
Company headquarters
LinkedIn lists San Francisco as headquarters, while Crunchbase lists Tel Aviv, Israel. source
Company legal name
Crunchbase identifies the legal entity as DagsHub Inc. source
Company size
LinkedIn lists DagsHub as privately held with 11–50 employees. source
Compute limitation
The documentation says DagsHub does not provide compute resources, so training and deployment run on infrastructure supplied by the customer. source
Data and code versioning
The product combines Git and DVC-based versioning so code and large data files can be managed together. source
Data Engine
Data Engine provides pandas-like curation, visualization, quality review, lineage, hard-example mining, labeling, auto-labeling, and data streaming. source
Data Engine
Data Engine provides APIs and tools for connecting sources, enriching metadata, querying, visualizing, annotating and producing dataloaders. source
Deployment options
Enterprise installations can run on-premises, in a VPC, in air-gapped environments, and in OpenShift environments. source
Documentation and support
DagsHub provides quick starts, feature guides, API reference, Python client reference, CLI reference, tutorials, FAQ material and Discord/community support. source
Enterprise security
DagsHub states that it adheres to ISO 27001 and ISO 9001 standards and supports custom OIDC and LDAP authentication for SSO. source
Experiment tracking
DagsHub supports experiment comparison, metrics and parameter visualization, real-time progress monitoring, and reproducibility. source
Experiment tracking
Experiments can be tracked with Git, DVC experiments or MLflow, including metrics, parameters and artifacts. source
Free-plan limits
The Individual plan allows up to 100 tracked experiments in private repositories, up to 2 private collaborators, and 20GB of DagsHub Storage. source
Funding
CB Insights reports $4.5 million raised across five rounds, including a $2.5 million seed round dated March 1, 2022. source
Generated API clients
DagsHub links auto-generated clients for multiple languages but warns that they have not been fully tested. source
GitHub synchronization
GitHub repositories can sync through webhooks, while GitHub pull requests and issues are shown in DagsHub and vice versa. source
Hosted scope
DagsHub hosts data but does not currently provide compute resources, relying on the user's local, cloud, or edge infrastructure. source
Independent API limitation
The same evaluation reported that MLflow run information stored in DagsHub could not then be retrieved programmatically through a Python API. source

Checked against the maker's pages on 15 Oct 2026.