Plans for every team
Start in the cloud, or run the same product on your own hardware.
SaaS Team
One team, hosted on AWS cloud
Sign up, allowances apply
Get startedUp to 10 members and 20 projects
1 TB storage included
100 GPU-hours per month for workflows and training
Every data, training and deployment feature
No servers to run, always up to date
Workstation
Multiple teams on a GPU workstation we deliver
Custom quote
Configure yoursA Dell workstation with GPUs and EmbodiFlow preinstalled
Unbox, plug in and start: nothing to install
Everything in SaaS Team, for multiple teams
Fully offline, data never leaves the building
Perpetual use, no license to renew
Backup, restore and guided upgrades
Enterprise
Multiple teams on your private cloud cluster
Custom quote
Configure yoursEverything in Workstation
Multi-node cluster with high availability
Scales GPU nodes with your workloads
Custom features and integrations
Dedicated engineer and on-site installation
SaaS Team allowances renew monthly. Need more? Contact us for additional capacity.
Compare plans
| Feature | SaaS Team | Workstation | Enterprise |
|---|---|---|---|
| Deployment and scale | |||
Where it runs | AWS cloud | Dell GPU workstation, delivered | Your private cloud cluster |
Teams | 1 | Unlimited | Unlimited |
Members | 10 | Unlimited | Unlimited |
Projects | 20 | Unlimited | Unlimited |
Storage | 1 TB | Your hardware | Your hardware |
Workflow CPU | 2,000 vCPU-hours / month | Your hardware | Your hardware |
GPU for workflows and training | 100 GPU-hours / month | Your hardware | Your hardware |
Concurrent training jobs | 2 | Your hardware | Your hardware |
Download traffic | 1 TB / month | Unlimited | Unlimited |
| Ingest and formats | |||
Browser and CLI bulk upload, resumable | |||
ROS bag, MCAP, vendor HDF5, LeRobot, UMI and egocentric headsets | |||
Import from Hugging Face Hub and S3-compatible storage | |||
Collection orders and delivery acceptance | |||
| Workflows and algorithms | |||
Visual workflow editor and reference templates | |||
Built-in standardizers, perception and post-processing | |||
Custom algorithms in Python or container images | |||
Team formats for private recorder formats | |||
GPU steps (depth, hand pose, SLAM, anonymization) | |||
| Quality, annotation and search | |||
Quality gates with built-in and custom checks | |||
Face anonymization and privacy admission | |||
Studio multi-stream replay | |||
AI-suggested episodes, subtasks and skills | |||
Annotation assignment and independent review | |||
Semantic search, similarity and data map | |||
| Versions, training and deployment | |||
Immutable dataset versions with lineage | |||
Export to LeRobot v3, RLDS, Zarr and HDF5 | |||
Hub-compatible reads for huggingface_hub and LeRobot | |||
Hosted training for policy and VLA models | |||
Offline evaluation against a baseline | |||
Inference services and edge deployment packages | |||
| Developers and governance | |||
REST API, TypeScript and Python SDKs, CLI | |||
MCP server for AI assistants | |||
Scoped personal access tokens and webhooks | |||
Team and project roles, audit logs | |||
Data stays in your own environment | |||
Perpetual use, no license to renew | |||
Backup and point-in-time restore | |||
Custom features and integrations | |||
| Support | |||
Email support | |||
Installation | Preinstalled | On-site | |
Dedicated engineer | |||
Frequently asked questions
It uses our AWS storage and GPUs. Workstation and Enterprise run on your hardware, limited only by it.
Yes. Workstation and Enterprise also run fully offline.
A Dell workstation with NVIDIA GPUs and EmbodiFlow preinstalled. Plug it in and start. Size disks and GPUs on the Customize page.
No. Both are yours to keep, with no license to renew. Your data stays on your hardware.
You do. Exports use open formats any tool can read.