canonical: https://jentic.com/apis/labelbox.com/labelbox

# Labelbox Foundry API

Jentic publishes the only available OpenAPI specification for Labelbox Foundry API, keeping it validated and agent-ready. Labelbox Foundry exposes packaged AI/ML models for image, text, and document tasks. The three-endpoint API lets clients create prediction jobs against a Foundry app, retrieve prediction results by ID, and search prediction jobs run against a given app. Authentication is HTTP bearer with a Labelbox-issued API key.

## For AI agents

Run Labelbox Foundry model predictions: create a prediction job for a Foundry app, fetch results by ID, and search prior prediction jobs.

## Scope

Does not handle data labelling jobs, dataset management, or training pipelines - use for Labelbox Foundry packaged-model inference only.

## Capabilities

- Create a prediction job for a Foundry app via POST `/foundry-app/{appId}/predict` with the input data and run parameters
- Retrieve completed prediction results by job ID via GET `/foundry-app/prediction/{id}`
- Search and filter prior prediction jobs for a Foundry app via POST `/foundry-app/{appId}/predictions`
- Run packaged Foundry models without managing model hosting or inference infrastructure
- Authenticate every call with a bearer token issued from the Labelbox dashboard

## Use cases

### Image Classification at Scale

Run a packaged Foundry vision model over a batch of images. POST `/foundry-app/{appId}/predict` creates the job with input data and parameters; GET `/foundry-app/prediction/{id}` polls for completion and returns the predictions. This avoids hosting and operating the model directly while still benefitting from a stable inference contract.

Example prompt: POST `/foundry-app/{appId}/predict` with the image batch, then poll GET `/foundry-app/prediction/{id}` until status is complete and read the labels.

### Document Extraction Workflow

Use a Foundry document-understanding app to extract structured fields from PDFs. The agent submits each document to POST `/foundry-app/{appId}/predict`, polls GET `/foundry-app/prediction/{id}` for the structured output, and writes the resulting fields into downstream systems. POST `/foundry-app/{appId}/predictions` can search past runs to avoid duplicate processing.

Example prompt: POST `/foundry-app/{appId}/predict` with the PDF reference, poll GET `/foundry-app/prediction/{id}`, and write the extracted fields into the data warehouse.

### Audit Trail of Past Predictions

Maintain an audit trail of model predictions for compliance and debugging. POST `/foundry-app/{appId}/predictions` searches prior prediction jobs for the app, returning IDs that can be fetched in detail via GET `/foundry-app/prediction/{id}.` This supports investigation when downstream systems flag anomalies.

Example prompt: POST `/foundry-app/{appId}/predictions` filtered to the past 24 hours and emit a list of recent run IDs into the audit log.

### AI Agent ML Operator

An agent in an ML-ops chat client triggers prediction jobs, polls results, and surfaces anomalies on instruction. Through Jentic, the agent calls Labelbox Foundry by intent without storing the bearer token in the agent runtime.

Example prompt: On 'classify these 200 images with the product-category model', POST `/foundry-app/{appId}/predict` with the image references and report back the IDs and statuses.

## Key endpoints

| Method | Path | Description |
| --- | --- | --- |
| POST | `/foundry-app/{appId}/predict` | Create a prediction job |
| GET | `/foundry-app/prediction/{id}` | Get prediction results |
| POST | `/foundry-app/{appId}/predictions` | Search prediction jobs |

## Key resources

- **Predictions** — Create, retrieve, and search prediction jobs for Foundry apps.

## Why Jentic

- **Setup:** Wiring the Labelbox Foundry API by hand means learning its bearer auth, targeting the app.labelbox.com host, and polling for async prediction results yourself. Through Jentic you install once, import the Labelbox Foundry API from the API Directory, store the bearer token once, and your agent calls it.
- **Permission scoping:** Labelbox Foundry puts the app id in the URL path (`/foundry-app/{appId}/predict`), so a rule can pin your agent to one Foundry app: it can run predictions and read their results for that app and nothing else. You choose the operations it may call, so submitting a batch prediction is included only if you add it.
- **Credential handling:** Your Labelbox bearer token is stored once, encrypted, by your own Jentic One instance and injected at execution time. It never enters the agent's prompt, logs, or context.
- **Discovery method:** Agents search Jentic by intent such as 'run a Labelbox Foundry prediction' or 'get a Foundry prediction result', and Jentic returns the matching Labelbox Foundry operation with its input schema so the agent calls the right endpoint without browsing the reference docs.

## Related APIs

- **Scale AI API** — Data labelling and model inference platform competing with Labelbox.
- **OpenAI API** — Foundation model API used alongside Foundry for general LLM tasks.
- **Anthropic Messages API** — Claude messages API for general LLM reasoning over Foundry outputs.
- **Clarifai API** — Hosted vision and text models with similar packaged-inference scope.

## FAQ

### Why is there no official OpenAPI spec for Labelbox Foundry API?

Labelbox documents Foundry on docs.labelbox.com but does not publish a canonical OpenAPI specification. Jentic generates and maintains this spec so that AI agents and developers can call Labelbox Foundry API via structured tooling. It is validated against the live API and kept up to date. Get started with Jentic One, the self-hosted execution layer.

### What authentication does the Labelbox Foundry API use?

The Foundry API uses HTTP bearer authentication: every request needs an Authorization header with a Labelbox-issued API key. Through Jentic, the key is stored encrypted in your Jentic One instance and injected per request.

### Can I run a prediction job with the Labelbox Foundry API?

Yes. POST `/foundry-app/{appId}/predict` creates a prediction job for a Foundry app with the input data and run parameters; GET `/foundry-app/prediction/{id}` returns the result once the job completes.

### How do I retrieve prediction results through Jentic?

Search Jentic for 'get Labelbox Foundry prediction results'. Jentic returns GET `/foundry-app/prediction/{id}` with its schema. Provide the prediction ID and Jentic executes the call using your stored bearer token.

### What rate limits apply to the Labelbox Foundry API?

Per-account rate limits aren't exposed in the public spec; Foundry pricing and concurrency are tied to your Labelbox plan. Refer to docs.labelbox.com or your Labelbox account team for current ceilings before scaling agent workloads.

### Can I search prior Foundry prediction jobs?

Yes. POST `/foundry-app/{appId}/predictions` searches prior prediction jobs for a given Foundry app, which is useful for building audit trails, deduplicating runs, and debugging anomalous results.

### Can I limit what my agent is allowed to do with the Labelbox Foundry API?

Yes. Because you run Jentic One yourself, your own rules decide which Labelbox Foundry operations and credentials the agent may use. Since the Foundry app id sits in the URL path, such as POST `/foundry-app/{appId}/predict`, you can pin the agent to a single app so it only runs predictions and reads their results via GET `/foundry-app/prediction/{id}` for that app. You also choose which operations are exposed, so searching prior jobs through POST `/foundry-app/{appId}/predictions` is available only if you add it.