Skip to content
NAMEFRAMECommercial PreviewApply for Pilot
Apply for Pilot
Developers

Everything the Studio does, without the Studio.

The editor panel is a convenience, not the product boundary. The same capture, labelling, validation and export path is available as a command line, a Python package, a REST API and an MCP server, and all four write the same evidence to the same places.

CLI commands
110
MCP tools
38
Read-only tools
24
Approval-gated
10

Four surfaces, one pipeline.

Pick whichever fits how you work. Nothing is exclusive to one of them.

CLI

110 commands

Capture from a running editor, label a dump, export a dataset, validate it, compare two runs, evaluate a model against one, and package a release. The full list ships with the package and is generated from the parser itself.

Python

The SDK

The same functions the CLI calls, importable. Useful when a run is one step inside a larger training pipeline rather than a thing you do by hand.

REST

API v1

Jobs, previews, tasks and analytics reports over HTTP, with authentication and role checks. Run it locally with a single command.

  • capture-unreal
  • scene-unreal
  • scene-validate
  • generation-plan
  • studio-dry-run
  • studio-validate
  • label
  • generate
  • run
  • verify
  • report
  • dataset
  • validate
  • dataset-inspect
  • distribution-preview
  • coverage-eval
  • domain-gap
  • model-eval
  • failure-mine
  • hard-cases
  • live-smoke
  • live-soak
  • api-serve
  • mcp-serve

Start a run from a terminal.

A capture against a live editor, labelled and exported, with the report written beside it.

# capture from a running editor over Remote Control; the job file names the dump
nameframe capture-unreal job.json --remote-url http://127.0.0.1:30010

# derive labels from the identity buffer, check them, then write the report
nameframe label  _out/run1/dump _out/run1/labels
nameframe verify _out/run1/dump
nameframe report _out/run1/dump _out/run1/report

# export and grade
nameframe dataset  _out/run1/dump _out/run1/dataset --format yolo,coco
nameframe validate _out/run1/dataset --out _out/run1/validation

# or all of it in one call
nameframe capture-unreal job.json --remote-url http://127.0.0.1:30010 --verify \
  --label-out _out/run1/labels --report-out _out/run1/report \
  --dataset-out _out/run1/dataset --dataset-format yolo,coco \
  --validate-out _out/run1/validation

Let agents inspect, plan and operate the data workflow.

An agent is good at the part where you describe what you want and something has to turn it into a scenario. It is a bad choice for the part that spends an hour of GPU time without asking.

Illustration of the protocol
  1. userInspect this scene and prepare a 1,000-frame pedestrian dataset.
  2. toolnameframe.scene.scanread-only
  3. agentScene inspected: 4 spawn zones, 2 camera zones, 6 classes.
  4. toolnameframe.scenario.validateread-only
  5. agentScenario drafted. Validation passed.
  6. toolnameframe.scenario.dry_runread-only
  7. agentDry run planned 1,000 placements, 0 refused.
  8. toolnameframe.preview.generatewrites, R1
  9. agent8-frame preview generated.
  10. toolnameframe.dataset.generateapproval required
  11. gateWaiting for approval.

The last step does not run. An agent can inspect a scene, draft a scenario, validate it and plan every placement without permission, because none of that changes anything. Starting the run is a separate decision that a person makes.

R0Inspection only21 tools
  • nameframe.server.health
  • nameframe.enterprise.status
  • nameframe.observability.status
  • nameframe.scene.get
  • nameframe.scene.scan
  • nameframe.scene.validate
  • nameframe.class.list
  • nameframe.run.get
  • nameframe.analytics.summary
  • nameframe.metadata.search_frames
  • nameframe.analytics.compare_runs
  • nameframe.analytics.annotation_issues
  • nameframe.model.evaluate
  • nameframe.model.failures
  • nameframe.model.compare_synthetic_real
  • nameframe.experiment.regression_campaign
  • nameframe.scenario.validate
  • nameframe.coverage.evaluate
  • nameframe.coverage.gaps
  • nameframe.task.list
  • nameframe.task.get
R1Cheap and reversible7 tools
  • nameframe.model.hard_case_plan
  • nameframe.model.domain_gap
  • nameframe.scenario.dry_run
  • nameframe.preview.generate
  • nameframe.dataset.plan
  • nameframe.coverage.correction_plan
  • nameframe.dataset.validate
R2Writes output or spends machine time6 tools
  • nameframe.scenario.updateapproval
  • nameframe.spawner.updateapproval
  • nameframe.camera.updateapproval
  • nameframe.environment.updateapproval
  • nameframe.scene.helper.createapproval
  • nameframe.dataset.cancelapproval
R3Long or expensive run4 tools
  • nameframe.dataset.generateapproval
  • nameframe.dataset.retryapproval
  • nameframe.dataset.resumeapproval
  • nameframe.model.generate_hard_casesapproval