Software for robotics teams

Know your data.
Understand your robot.

Cupel turns recordings into reviewed datasets.
Tether Studio helps you test and improve robot policies.

Interactive product tour
Tether Studio /PushT evaluation
Recorded simulation
MATCHED COMPARISON

Did the change help?

Same seeds · 101, 102
BaselineAction chunk 8
Recorded PushT baseline: the block aligned with its target at the end of episode oneEpisode 1
Successful episodes1 / 2
CandidateAction chunk 4
Recorded PushT candidate: the block remained short of the target at the end of episode oneEpisode 1
Successful episodes0 / 2
6 / 6

Real saved frames. Sampled replay; the final frame is each policy’s episode end.

Replay a recorded episode and inspect what the policy actually did.Explore Tether Studio

A guided look at the workflows, not a live application. Product availability varies by use case.

Find your starting point

Two workspaces.
Different jobs to do.

Start with the problem in front of you.
You can use either product on its own.

01 / DATA

Cupel

“Can I use this data?”

Inspect recordings, understand data problems, and review the changes that turn raw sources into a dataset you can use.

  • Find issues in the original recording
  • Compare source and derived data
  • Keep lineage with each dataset release
Meet Cupel

Robot datasets · Video · Spatial data workflows

02 / POLICIES

Tether Studio

“Is this policy actually better?”

Watch a policy run, investigate a failure, and compare a candidate using the same experiment settings.

  • Start in simulation without a robot
  • Replay observations, actions, and timing
  • See regressions before choosing a version
Meet Tether Studio

Simulation · Investigation · Policy comparison

Built for the review, too

The result matters.
So does how you got there.

01 / SOURCE

Keep the original.

A derived dataset or a candidate policy should not erase the baseline you need to compare it with.

02 / EVIDENCE

Show what changed.

Open the recording, inspect the selected event, and follow the evidence behind a decision.

03 / LIMITS

Make unknowns visible.

Missing observations, untested targets, and regressions should be part of the review.

Before you start

A few practical questions.

Do I need a physical robot?

No. Tether Studio has local simulation workflows. Cupel starts with your recordings or datasets. Hardware integration is a separate step, not a requirement for exploring the products.

Do Cupel and Tether Studio have to work together?

No. Cupel focuses on data preparation and review. Tether Studio focuses on policy execution, investigation, and comparison. Choose the one that fits your current task.

What is available today?

Tether Studio supports local simulation, recording inspection, matched comparisons, and supported simulation packages. Cupel has dataset inspection, treatment, evaluation, and release workflows, with video and spatial capabilities at different stages. The product pages describe the current scope; contact us about your specific inputs and target.

How do I get started?

Email team@fastcrest.com with the dataset or policy you are working with and what you want to improve. We will discuss fit, access, and the next step for your workflow.

Build with FastCrest

Bring a dataset.
Or a policy.

Tell us what you are working on.
We’ll help you find the right starting point.

Talk to the team