Tether Studio · Robot policy development

Understand what your robot policy does.

Run a policy, inspect a failure, and test a candidate without rebuilding the experiment. Start entirely in simulation, with no physical robot required.

Talk to us about Tether Studio Explore the workflow

Tether Studio /PushT evaluation

Recorded simulation

MATCHED COMPARISON

Did the change help?

Same seeds · 101, 102

Recorded PushT baseline: the block aligned with its target at the end of episode one — **Baseline**Action chunk 8

Recorded PushT candidate: the block remained short of the target at the end of episode one — **Candidate**Action chunk 4

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.Discuss your policy

Condensed workflow preview. Not a live application.

Your work in Tether Studio

Observe → Investigate → Change → Compare → Decide

Run and observe

Use a prepared simulation workspace to run a controller or a supported trained policy. Keep the actual configuration, task settings, and recording.

Inspect and investigate

Replay observations and actions, select a moment or segment, and inspect timing and logs. Keep observed facts separate from possible explanations.

Compare and decide

Test a candidate with matching task settings and seeds. Review outcomes, timing, and regressions side by side, then keep the baseline or candidate.

Preserve the result

Return to saved runs and investigations, retain comparison decisions, and prepare supported simulation packages with their evidence.

Start with your use case.

Tether Studio currently focuses on local simulation workflows. Hardware connections, fleet deployment, and broader training workflows are still being developed. Simulation results are not a claim of readiness on a physical robot.

Tell us what you are working on so we can discuss the workflows that fit.

Contact FastCrest