Cloud list prices, one JSON call away.

Reading the numbers

What a price here means

List price only

No Savings Plan, Reserved Instance, Committed Use or negotiated discount — unless you describe your commitments with /api/v1/price-request, or ask /compare for a commitment_term.

Usage is priced when you give it

Pass quantities like GB stored, requests or GB processed in usage; anything usage-based you leave out is named in the result's unpriced list, with the usage key that would price it.

Free tiers aren't deducted

They apply per account, not per resource.

Monthly = hourly × 730

The providers' own convention, unless you pass usage.hours for something that doesn't run all month. Prices are in USD.

One vocabulary for all three clouds

size, os, count, purchase_model and the rest mean the same thing on every type — see normalized attributes. A provider's own record works too, as raw.

Every answer says which price list it came from

Each priced resource carries price_list_version and fetched_at.

Why free

Built because we needed it

We built this because Lensix needed it. Lensix scans AWS, Azure, and GCP accounts and has to know what every resource actually costs across clouds, and there was no clean source for that. So we built one, and it turned out everyone else needs it too. Cloudcostdata.com is maintained by Absolute Ops and runs the same pricing data Lensix uses every day.

Access

Get an API key

Create a free account, verify your email, and create keys from your dashboard — as many as you need, one per app or environment, each revocable on its own. Keyed calls are limited to 60 a minute per account, /compare to 10 of those; every response says how many remain.

Reference

API

Base URL is https://api.cloudcostdata.com. Send your key in the X-API-Key header. Request and response bodies are JSON.

Price one resource, or up to 500 in {"resources": [...]}. Each resource takes a resource_type named for its cloud (aws.ec2_instance, azure.vm, gcp.compute_instance), a region, an optional resource_id, its attributes in the normalized vocabulary (count is the multiplier for identical resources), and optionally usage: monthly quantities for usage-based charges (see resource types for what each takes). A provider's own record shape (the boto3, Azure SDK or Compute Engine REST response) also works, as raw in place of attributes. A batch may add as_of and commitments with the same meaning as in /price-request.

curl -s https://api.cloudcostdata.com/api/v1/estimate \
  -H "X-API-Key: $CCD_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "resources": [
      {"resource_type": "aws.ec2_instance", "region": "us-east-1",
       "attributes": {"size": "m5.large", "count": 3}},
      {"resource_type": "aws.s3_bucket", "region": "us-east-1",
       "usage": {"storage_gb": 1000, "read_requests": 10000000}},
      {"resource_type": "azure.vm", "region": "eastus",
       "attributes": {"size": "Standard_D2s_v3", "os": "linux"}},
      {"resource_type": "gcp.compute_instance", "region": "us-central1",
       "attributes": {"size": "n2-standard-4", "purchase_model": "spot"}}
    ]
  }'

Response (abridged):

{
  "currency": "USD",
  "hours_per_month": 730,
  "total_monthly": 506.88,
  "total_hourly": 0.694356,
  "resources": [
    {
      "resource_type": "aws.ec2_instance", "provider": "aws", "region": "us-east-1",
      "count": 3, "purchase_model": "on_demand", "monthly": 210.24, "hourly": 0.288,
      "components": [{"label": "m5.large (Linux, Shared)", "price_per_unit": 0.096,
                      "quantity": 2190, "unit": "Hrs", "monthly": 210.24}],
      "unpriced": [{"usage": null, "label": "data transfer"}],
      "notes": ["attached EBS volumes are separate ebs_volume resources ..."],
      "basis": "AWS published list price, 730 hours/month; On-Demand unless ...",
      "price_list_version": "20260903201523", "fetched_at": "2026-10-29T02:10:44+00:00"
    },
    {"resource_id": null, "resource_type": "gcp.compute_instance", "region": "us-central1",
     "error": "no published GCP SKU matched: ..."},
    ...
  ]
}

A resource that's valid but can't be priced — an unknown type, or a region we haven't loaded — comes back in its position in resources as {"resource_id", "resource_type", "region", "error"}, and the rest of the batch is still priced. A malformed body — an unknown attribute, an enum value that isn't one of the documented ones, a usage key the type doesn't take, purchase_model disagreeing with commitment_term — is a 400 for the whole call.

Describe what you need once and get each cloud's cheapest matching options in the regions you name, priced exactly as /estimate would price them. Categories: vm, block_storage, object_storage, data_transfer_out, kubernetes_control_plane, serverless_functions, relational_database and nat_gateway; GET /api/v1/compare/categories lists the requirements each takes. Leave a cloud out of regions to skip it, or send "region_set": "us" (eu, asia) instead and read the region chosen per provider from the result. For vm and relational_database, "commitment_term": "1yr" or "3yr" compares committed prices instead: AWS Savings Plans and Reserved Instances, Azure reservations, GCP committed use discounts; "purchase_model": "spot" compares the Spot rate on each cloud, and architecture (x86_64, arm64, any) narrows the shapes.

curl -s https://api.cloudcostdata.com/api/v1/compare \
  -H "X-API-Key: $CCD_API_KEY" -H "Content-Type: application/json" \
  -d '{
    "category": "vm",
    "regions": {"aws": "us-east-1", "azure": "eastus", "gcp": "us-central1"},
    "requirements": {"vcpus": 4, "memory_gib": 16, "os": "linux"},
    "limit": 3
  }'

Response (abridged): ranking is each cloud's cheapest option, cheapest first; results has every option with its spec and full estimate.

{
  "category": "vm",
  "ranking": [
    {"provider": "aws", "region": "us-east-1", "resource_type": "aws.ec2_instance",
     "spec": {"size": "m6g.xlarge", "vcpus": 4, "memory_gib": 16, "architecture": "arm64",
              "purchase_model": "on_demand"},
     "monthly": 112.42},
    ...
  ],
  "results": {
    "aws": {"region": "us-east-1", "options": [{"resource_type": "aws.ec2_instance",
            "attributes": {"size": "m6g.xlarge", ...}, "spec": {...},
            "monthly": 112.42, "hourly": 0.154, "estimate": {...}}, ...], "errors": []},
    ...
  }
}

The full cost-request document: resources plus your AWS Reserved Instances and Savings Plans, applied in AWS's own order, returning effective cost, coverage and unused commitment.

curl -s https://api.cloudcostdata.com/api/v1/price-request \
  -H "X-API-Key: $CCD_API_KEY" -H "Content-Type: application/json" \
  -d '{
    "format": "cost-request", "format_version": 2,
    "as_of": "2026-10-01T00:00:00Z",
    "resources": [{"resource_id": "web", "resource_type": "aws.ec2_instance",
                   "region": "us-east-1",
                   "attributes": {"size": "r5.xlarge", "count": 4}}],
    "commitments": [{"commitment_id": "sp-1", "commitment_type": "savings_plan",
                     "plan_type": "compute", "commitment_hourly": 0.5,
                     "term_years": 1, "payment_option": "No Upfront"}]
  }'

The call behind the widget at the top of this page: a VM comparison with a fixed shape — vcpus (1–64), memory_gib (0.5–512), os, commitment_term and a region_set of us, eu or asia — answered exactly as /compare would, three options per cloud. No key; 30 calls per 15 minutes per address, after which a 429 with Retry-After. For any other shape, region or category, use /compare with a key.

Every priceable resource type, its provider, the attributes it takes and the usage it can price — the table below, as JSON.

Which provider/service/region price lists are loaded, at which published version and when they were fetched. A region missing here can't be priced yet.

Which account and key is calling, with that key's request count and last use.

Errors

Vocabulary

Normalized attributes

The same concept has one name on every type, whatever the cloud calls it. Where you'd rather use the provider's own field names, send the provider's own record as raw instead.

size

The provider's compute shape: m5.large, Standard_D2s_v3, n2-standard-4, db.r5.large, db-custom-2-7680, a WorkSpaces bundle, an App Service plan size.

count

Positive integer, default 1: the only multiplier. Nodes in a cluster, brokers, shards, scale-set instances, or simply how many identical resources this entry describes.

os

linux, windows, rhel, suse, ubuntu_pro.

state

running or stopped; a stopped resource has no compute charge.

attached

Boolean, for IP addresses and volumes.

purchase_model

on_demand (default), spot where the cloud sells it, or committed, which is required with — and implied by — commitment_term 1yr or 3yr. Spot on AWS prices through the public Spot feed.

Units are in the name

_gb gigabytes, _gib gibibytes, _mbps MB/s, _mibps MiB/s, _tb / _tib likewise. Memory is memory_gib on VM-class types and memory_mb on function-class ones.

Attributes vs usage

attributes are provisioned or configured values; usage is metered consumption per month. Every hourly-billed type takes usage.hours, default 730.

Enums

Documented values are lowercase snake_case. Send them in any casing with -, _ or spaces, or in the provider's own spelling (Standard_LRS, Linux/UNIX, ZoneRedundant), and they come back canonical.