ML Infrastructure Bearer Token

Anyscale REST API

Scalable ML infrastructure for Ray workloads

Anyscale provides a unified platform for building, deploying, and managing distributed machine learning applications using Ray. The API enables developers to programmatically create compute clusters, deploy ML models, manage workloads, and monitor distributed training jobs. It's designed for teams scaling from prototypes to production ML systems with complex distributed computing requirements.

Base URL https://api.anyscale.com/v2

API Endpoints

Method	Endpoint	Description
GET	`/clusters`	List all Ray clusters in your organization, returning cluster IDs, states, and compute configurations.
POST	`/clusters`	Create a new Ray cluster with specified compute resources
GET	`/clusters/{cluster_id}`	Get details about a specific Ray cluster including node count, runtime version, and current state.
DELETE	`/clusters/{cluster_id}`	Terminate a running Ray cluster, requiring cluster_id in the path.
POST	`/services`	Deploy a Ray Serve application as a managed service
GET	`/services`	List all deployed Ray Serve services with their endpoints and health status.
GET	`/services/{service_id}`	Get details and status of a deployed service including replicas, routes, and deployment config.
PATCH	`/services/{service_id}`	Update a deployed service configuration such as replicas, autoscaling, or resource limits.
DELETE	`/services/{service_id}`	Delete a deployed Ray Serve service, requiring service_id in the path.
POST	`/jobs`	Submit a Ray job for execution on a cluster, returning the job_id for tracking.
GET	`/jobs`	List all jobs and their execution status including pending, running, and completed states.
GET	`/jobs/{job_id}`	Get detailed information about a specific job including start time, duration, and exit code.
GET	`/jobs/{job_id}/logs`	Retrieve logs from a running or completed job
DELETE	`/jobs/{job_id}`	Cancel a running job, requiring job_id in the path and returning cancellation status.
GET	`/compute-configs`	List available compute configurations and instance types for cluster creation.

Code Examples

curl -X POST https://api.anyscale.com/v2/clusters \
  -H 'Authorization: Bearer anyscale_your_api_token_here' \
  -H 'Content-Type: application/json' \
  -d '{
    "name": "ml-training-cluster",
    "compute_config": "ml.m5.4xlarge",
    "min_workers": 2,
    "max_workers": 10,
    "ray_version": "2.9.0"
  }'

const response = await fetch('https://api.anyscale.com/v2/clusters', {
  method: 'POST',
  headers: {
    'Authorization': 'Bearer anyscale_your_api_token_here',
    'Content-Type': 'application/json'
  },
  body: JSON.stringify({
    name: 'ml-training-cluster',
    compute_config: 'ml.m5.4xlarge',
    min_workers: 2,
    max_workers: 10,
    ray_version: '2.9.0'
  })
});

const cluster = await response.json();
console.log('Cluster ID:', cluster.id);

import requests

headers = {
    'Authorization': 'Bearer anyscale_your_api_token_here',
    'Content-Type': 'application/json'
}

data = {
    'name': 'ml-training-cluster',
    'compute_config': 'ml.m5.4xlarge',
    'min_workers': 2,
    'max_workers': 10,
    'ray_version': '2.9.0'
}

response = requests.post(
    'https://api.anyscale.com/v2/clusters',
    headers=headers,
    json=data
)

cluster = response.json()
print(f"Cluster ID: {cluster['id']}")

Use Anyscale from Claude / Cursor / ChatGPT

Get a hosted MCP endpoint for Anyscale. Paste your Anyscale API key, copy back one URL, drop it into Claude Desktop, Cursor, or any AI client that supports remote MCP. Your AI calls Anyscale directly with your credentials — no local install, works on mobile.

create_ray_cluster Create and configure a new Ray cluster with specified compute resources and autoscaling parameters for distributed ML workloads

deploy_ml_service Deploy a trained ML model as a scalable Ray Serve service with automatic load balancing and version management

submit_training_job Submit a distributed training job to an Anyscale cluster with resource specifications and monitoring capabilities

monitor_cluster_metrics Retrieve real-time metrics and resource utilization data from Ray clusters including CPU, GPU, and memory usage

manage_job_lifecycle Start, stop, monitor, and retrieve logs from Ray jobs running on Anyscale infrastructure

Connect in 60 seconds

Paste your Anyscale key → get an MCP URL → paste into Claude/Cursor. Hosted by IOX, encrypted at rest.

Connect Anyscale to your AI →

Anyscale REST API

API Endpoints

Sponsor this page

Code Examples

Use Anyscale from Claude / Cursor / ChatGPT

Connect in 60 seconds

Related APIs