Headless workflows and API reference

Choose the right integration

End-to-end REST path

1. Create an account and API key

   Complete account, organization, and key setup in the platform. Copy the raw key once and store it in a secret manager.

2. Verify authentication

   Call /datasets/datasets/ with Authorization: Api-Key <raw_api_key>.

3. Create and verify a dataset

   Create a dataset, upload or import samples, then inspect labels, annotations, and split counts.

4. Create and run a task query

   Select the dataset, plan the pipeline, start a task run, and poll durable REST state. WebSockets are optional live updates.

5. Deploy a model version

   Select a completed version and checkpoint, create a deployment, call its predictor, and inspect usage and artifacts.

Follow the headless journey

Start with secure authentication, then move through data, model, serving, artifact, and agent workflows. Every guide links to the canonical production reference.

1. 01Quickstart and authenticationCreate an operational API key, verify it safely, and make your first server-side request.
2. 02Datasets and annotationsCreate datasets, ingest samples, import annotations, verify readiness, and preserve split integrity.
3. 03Pipelines, training, and resultsPlan a pipeline, start one run, follow durable state, recover checkpoints, and interpret completed results.
4. 04Deployments and predictionsCreate a deployment from an eligible model version, call its predictor, and inspect usage and result media.
5. 05Files, artifacts, and errorsRetrieve protected outputs correctly and handle endpoint-specific errors without unsafe assumptions.
6. 06Local MCP adapterConfigure the trusted local stdio adapter and use its production-backed tools without widening the trust boundary.
7. 07REST and MCP referenceUse the complete production-aligned endpoint and tool cards, coverage boundaries, and version history.

Return to the platform deployment guide