EmbodiFlowDocs
Get startedDataWorkflows and algorithmsTraining and deploymentDeveloper reference
  • EmbodiFlow documentation

Get started

  • Quickstart
  • Key concepts
  • Console tour

Team

  • Team and members
  • Roles and permissions

Help

  • FAQ

© 2026 EmbodiFlow

llms.txtOpenAPI JSON

Get started

EmbodiFlow documentation

EmbodiFlow turns robot and human recordings into datasets you can train on. Upload, process, review, version, export and train, all in one place.

EmbodiFlow is where robot teleoperation and first-person human recordings become training data. You bring the recordings. The platform turns them into reviewed episodes and frozen dataset versions that you can export or train on.

How it works

You create a project for one kind of data, robot or human, and upload recordings straight from the browser. From there the project's workflow takes over: it standardizes every recording to LeRobot v3, checks its quality, anonymizes faces and cuts it into episodes.

Then it's your turn. Look through the episodes and their quality results, keep the ones you trust, and freeze them into a dataset version. A version never changes, so when you export it as LeRobot v3 or train a model on it, you always know exactly what went in.

New here? The quickstart takes you through the whole journey in one sitting, and key concepts explains the words you'll meet on the way.

Get started

Quickstart

Take one project from the first upload to a LeRobot v3 dataset you can download.

Key concepts

The handful of words you need to use EmbodiFlow, and how they connect.

Console tour

Find your way around the console: the sidebar, the project tree, the account menu, Home, My work and Usage & billing.

Explore the docs

Data

Bring recordings in, watch them, review the episodes and their quality, then freeze a dataset version and export it.

Workflows and algorithms

Every upload runs through a workflow you can see and change. Use the built-in algorithms, or write your own in Python.

Training and deployment

Train a policy on a frozen dataset version, watch it learn, and take the checkpoint to your robot.

Developer reference

Authenticate, call the REST API, receive webhooks, and script EmbodiFlow from the CLI and SDK.

Team

Team and members

Invite people, change their roles, and find everything that belongs to the team: keys, storage, audit log, webhooks and usage.

Roles and permissions

What each team role and project role can do, and who can open raw recordings.