AWS Amazon SageMaker
Browse all Amazonaws APIs★ Only Publicly Available OpenAPI DocumentAI/MLMl Inferencehmac302 EndpointsREST
For Agents
Train, deploy, monitor, and govern machine-learning models end-to-end with 300+ operations covering training jobs, endpoints, pipelines, and feature stores.
Use for: I need to launch a training job for an XGBoost model on data in S3, Deploy a registered model to a real-time inference endpoint, Run batch inference on a large dataset using an existing model, Start a hyperparameter tuning job with 50 trials
Not supported: Does not invoke deployed endpoints for inference, run model code on edge devices, or send predictions to human reviewers — use for the SageMaker control plane (training, model registry, endpoint management, pipelines, feature store) only.
Jentic publishes the only available OpenAPI specification for Amazon SageMaker, keeping it validated and agent-ready. Amazon SageMaker is AWS's end-to-end machine-learning platform — it provides training jobs, hyperparameter tuning, model packaging, real-time and batch inference endpoints, feature stores, data labelling, AutoML, MLflow-style experiment tracking, model monitoring, and pipeline orchestration. The control-plane API exposed here covers all 300+ operations across the ML lifecycle, from launching a training job to deploying a multi-model endpoint behind production traffic.
Install Jentic One Beta
Connect the Amazon SageMaker to your agent
Jentic One is a self-hosted execution layer for AI agents. It lets your agent call the Amazon SageMaker, or any other public or private API you need. You set the rules, the agent never sees your credentials, and every call is logged.
Two steps, two machines. Install the instance in a safe environment, then register your agent from wherever it runs.
1
Step 1: Jentic One Host machine
# On the machine that will host your Jentic One instance:
curl -fsSL https://raw.githubusercontent.com/jentic/jentic-one/main/tools/install.sh | sh2
Step 2: Agent machine
# On the machine where your agent runs (keep this separate from the instance):
curl -fsSL https://raw.githubusercontent.com/jentic/jentic-one/main/tools/install.sh | sh
jentic register # connects your agent to your Jentic One instanceJentic One is in public beta. The setup above keeps your agent separate from the instance, which is what you want before using real credentials: an agent running as the same OS user as Jentic One can read its stored keys directly. Just evaluating? A single local install is fine to start. See the secure deployment guide for the tiers.
Capabilities
What an agent can do with Amazon SageMaker API.
Launch training and hyperparameter-tuning jobs with CreateTrainingJob and CreateHyperParameterTuningJob
Deploy real-time inference endpoints from registered model packages with CreateEndpoint and update them via UpdateEndpoint
Run batch inference with CreateTransformJob against large datasets in S3
Orchestrate ML pipelines and experiments with CreatePipeline, StartPipelineExecution, and Trial Components
Manage feature groups, model packages, and the SageMaker Model Registry for governance
Configure model monitoring, bias detection, and explainability jobs against deployed endpoints
Run AutoML jobs with CreateAutoMLJob and label data with CreateLabelingJob
Use Cases
Patterns agents use Amazon SageMaker API for, with concrete tasks.
★ Production model training and deployment
ML platform teams use SageMaker as the end-to-end pipeline for training and serving production models. CreateTrainingJob spins up managed training infrastructure with the team's container and S3 data; on completion, CreateModel registers the artifact and CreateEndpoint deploys it behind a managed inference endpoint with auto-scaling and multi-AZ failover. UpdateEndpoint handles zero-downtime model rollouts.
Call CreateTrainingJob with an XGBoost container, S3 input data, and an output path; on Completed status, register the artifact with CreateModel and deploy with CreateEndpoint to an ml.m5.large instance.
Hyperparameter tuning and AutoML
Data science teams use CreateHyperParameterTuningJob to run dozens of training trials in parallel across a defined search space, returning the best-performing model by objective metric. CreateAutoMLJob goes further — it explores algorithm choice and feature engineering automatically against tabular data. Both reduce the manual sweep work that typically slows model iteration.
Submit CreateHyperParameterTuningJob with a max of 50 trials, Bayesian strategy, and the objective metric set to validation:auc, then poll DescribeHyperParameterTuningJob until BestTrainingJob is reported.
Model governance with the Model Registry
ML governance teams use the SageMaker Model Registry (CreateModelPackageGroup, CreateModelPackage, UpdateModelPackage) to version models, attach approval status, and gate production rollouts. CI pipelines train new candidates and register them as PendingManualApproval; once approved, CreateEndpoint or UpdateEndpoint promotes the version to live traffic. The registry is the single source of truth for what is in production.
Call CreateModelPackage to register a new model version in group fraud-detector with ModelApprovalStatus=PendingManualApproval, then UpdateModelPackage to Approved once tests pass.
Agent-driven ML operations through Jentic
AI agents that automate ML lifecycle operations use Jentic to call CreateTrainingJob, CreateEndpoint, and StartPipelineExecution without holding raw AWS credentials. Jentic stores keys in the MAXsystem vault, signs each request with SigV4, and returns the structured response — so an agent can chain training, evaluation, registration, and deployment as a single workflow.
Search Jentic for 'launch a SageMaker training job', load CreateTrainingJob, execute it with the team's container URI and S3 paths, then on Completed call CreateModel and CreateEndpoint to deploy.
Key Endpoints
302 endpoints — jentic publishes the only available openapi specification for amazon sagemaker, keeping it validated and agent-ready.
METHOD
PATH
DESCRIPTION
POST
/
All SageMaker control-plane operations target the root path with X-Amz-Target headers — for example, SageMaker.CreateTrainingJob, SageMaker.CreateEndpoint, SageMaker.CreateModel, SageMaker.CreateTransformJob, SageMaker.CreateHyperParameterTuningJob, SageMaker.CreatePipeline, and SageMaker.StartPipelineExecution
POST
/
All SageMaker control-plane operations target the root path with X-Amz-Target headers — for example, SageMaker.CreateTrainingJob, SageMaker.CreateEndpoint, SageMaker.CreateModel, SageMaker.CreateTransformJob, SageMaker.CreateHyperParameterTuningJob, SageMaker.CreatePipeline, and SageMaker.StartPipelineExecution
Why Jentic?
Three things that make agents converge on Jentic-routed access.
Credential isolation
AWS credentials are stored encrypted in the Jentic vault (MAXsystem). Each SageMaker call is signed with SigV4 server-side and the agent only receives scoped, short-lived authorisation — raw access keys never enter the agent's context.
Intent-based discovery
Agents search by intent (e.g., 'launch a SageMaker training job') and Jentic returns matching SageMaker operations with their input schemas, so the agent calls CreateTrainingJob or CreateEndpoint without browsing the 300+ operation surface.
Time to first call
Direct SageMaker integration: 3-7 days for SigV4 signing, IAM execution-role plumbing, container packaging, and pagination across the 300+ control-plane operations. Through Jentic: under an hour for the API surface — search, load schema, execute (the IAM role and container still need to exist).
Related APIs
Alternatives and complements available in the Jentic catalogue.
FAQs
Specific to using Amazon SageMaker API through Jentic.
Why is there no official OpenAPI spec for Amazon SageMaker?
AWS does not publish an OpenAPI specification. Jentic generates and maintains this spec so that AI agents and developers can call Amazon SageMaker via structured tooling. It is validated against the live API and kept up to date. Get started at https://app.jentic.com/sign-up.
What authentication does the Amazon SageMaker API use?
SageMaker uses AWS Signature Version 4 HMAC signing with an access key ID and secret access key, optionally with a session token. SageMaker also requires an IAM execution role passed to most create-job operations so the service can read S3 and write logs on your behalf. Through Jentic, your AWS credentials live in the MAXsystem vault and Jentic signs each request server-side.
Can I run real-time and batch inference with the same Amazon SageMaker model?
Yes. Once a model is registered with CreateModel, you can deploy it behind a real-time endpoint with CreateEndpoint or run batch inference over S3 data with CreateTransformJob — both reference the same model artifact. The choice depends on latency requirements: endpoints for low-latency online traffic, transform jobs for offline scoring.
What are the rate limits for the Amazon SageMaker API?
AWS does not publish a fixed RPS for the SageMaker control plane in the spec; standard AWS API throttling applies and the service returns ThrottlingException on overage. There are also resource quotas (max concurrent training jobs, max endpoints) per region that return ResourceLimitExceeded — request increases via the AWS service quotas console.
How do I deploy a model to a real-time endpoint with Amazon SageMaker through Jentic?
Search Jentic for 'create a SageMaker endpoint', load the CreateModel, CreateEndpointConfig, and CreateEndpoint operations, and call them in sequence with your model artifact S3 URI, instance type, and execution role ARN. Poll DescribeEndpoint until EndpointStatus is InService. Install with pip install jentic.
Does the Amazon SageMaker API serve inference traffic directly?
No — this control-plane API creates and manages endpoints but does not invoke them. Inference requests go to a separate dataplane service (sagemaker-runtime) at runtime.sagemaker.{region}.amazonaws.com. Use this API to set up, scale, and monitor endpoints; use sagemaker-runtime to call InvokeEndpoint for actual predictions.
GET STARTED