AWS Extended Support Cost Calculator — MCP Server

Note: This tool calculates only the additional Extended Support charges. Standard usage costs (compute, storage, I/O, etc.) are separate and not included in these calculations.

What This MCP Does

This MCP server gives Kiro (IDE or CLI) and Amazon Quick the ability to answer questions and generate reports about AWS Extended Support across 5 services: RDS/Aurora, ElastiCache, EKS, OpenSearch, and DocumentDB.

You can ask Kiro things like:

• "Which EKS versions are currently in extended support?"
• "When does PostgreSQL 14 enter extended support?"
• "What's the extended support cost for a db.r6g.xlarge Aurora MySQL instance in us-east-1?"

And you can generate cost reports from CSV inventories:

• "Calculate extended support costs for /path/to/rds_inventory.csv and save the report to ~/Downloads"

The reports categorize your resources by support status (already in extended support, entering this year, entering next year) and project costs for the current and next calendar year based on exact extended support start dates.

Service	Pricing Model
RDS / Aurora	Per-vCPU-hour × Year tier (from AWS RDS Pricing API)
ElastiCache	On-demand + premium (80% Y1-2, 160% Y3) (from AWS ElastiCache Pricing API)
EKS	Additional per-cluster-hour rate (from AWS EKS Pricing API)
OpenSearch / Elasticsearch	Per NIH — Normalized Instance Hour (from AWS OpenSearch Pricing API)
DocumentDB	Per-vCPU-hour × Year tier (from AWS DocumentDB Pricing API)

Version lifecycle dates are fetched live from AWS APIs: RDS and EKS from their own APIs, and ElastiCache, OpenSearch, and DocumentDB from the AWS Health DescribeServiceLifecycle API, with built-in dates as a fallback. See How Calculations Work for details.

Installation

Prerequisites

• Python 3.10+ with uv installed
• An AWS profile with read-only access (see AWS Credentials below)
• Kiro IDE, Kiro CLI, or Amazon Quick

Install in Kiro

Add to your Kiro MCP config (~/.kiro/settings/mcp.json):

{
  "mcpServers": {
    "aws-extended-support": {
      "command": "uvx",
      "args": ["aws-extended-support-calculator@latest"],
      "env": {
        "AWS_PROFILE": "your-aws-profile"
      },
      "disabled": false,
      "autoApprove": [
        "rds_cost", "elasticache_cost", "eks_cost", "opensearch_cost", "documentdb_cost",
        "rds_report", "elasticache_report", "eks_report", "opensearch_report", "documentdb_report",
        "extended_support_report",
        "list_services", "list_versions", "list_billing_events", "check_data", "refresh_data"
      ]
    }
  }
}

Restart Kiro. The server is downloaded automatically on first use — no pip install needed.

Prerequisite: uv must be installed (brew install uv on Mac).

Install in Amazon Quick

In the Amazon Quick desktop app, go to Settings → Capabilities → Connectors → MCP Servers and add a server that runs the command uvx with the argument aws-extended-support-calculator@latest. Set the AWS_PROFILE environment variable to your AWS profile. No --kiro setup step is needed.

AWS Credentials

The MCP server uses an AWS profile for API calls. Make sure you have a profile configured with access to an AWS account.

If you don't have a profile yet, set one up:

aws configure --profile my-profile-name

This will prompt for your access key, secret key, and default region.

What permissions are needed?

The server needs read-only access to four APIs:

API Call	Purpose
ec2:DescribeInstanceTypes	Look up vCPU counts for instance types
rds:DescribeDBMajorEngineVersions	Fetch RDS/Aurora version lifecycle dates
eks:DescribeClusterVersions	Fetch EKS version lifecycle dates
health:DescribeServiceLifecycle	Fetch ElastiCache, OpenSearch, and DocumentDB version lifecycle dates

Minimum IAM policy:

{
  "Version": "2012-10-17",
  "Statement": [
    {
      "Effect": "Allow",
      "Action": [
        "ec2:DescribeInstanceTypes",
        "rds:DescribeDBMajorEngineVersions",
        "eks:DescribeClusterVersions",
        "health:DescribeServiceLifecycle"
      ],
      "Resource": "*"
    }
  ]
}

If the profile can't call health:DescribeServiceLifecycle, the server uses built-in dates for ElastiCache, OpenSearch, and DocumentDB. Ask "Is the pricing data fresh?" (check_data) to see which source is in use and why.

What works without credentials?

All pricing data is fetched from public AWS Pricing APIs — no credentials needed:

• RDS, ElastiCache, DocumentDB, EKS, and OpenSearch pricing
• Built-in fallback lifecycle dates for ElastiCache, OpenSearch, and DocumentDB

vCPU lookups and live version lifecycle dates (RDS, EKS, and the Health API for ElastiCache, OpenSearch, and DocumentDB) require credentials. Once cached (24-hour TTL), the server works without credentials until the cache expires.

CSV Format Requirements

The batch report tools auto-detect columns by matching header names (case-insensitive). Your CSV must include headers that match one of the recognized names below.

RDS / Aurora / DocumentDB

Field	Required	Recognized Column Names
Instance type	Yes	Instance Type, InstanceType, DBInstanceClass, class
Engine	Yes (RDS only)	DB Engine, Engine, DatabaseEngine
Version	Yes	Engine Version, Engine Full Version, EngineVersion, Version
Region	Yes	Instance Region, Region, RegionCode
Instance name	No	Instance Name, DBInstanceId, Identifier, Label
Multi-AZ	No	Multi-AZ, MAZ Flag, MultiAZ

ElastiCache

Field	Required	Recognized Column Names
Node type	Yes	Node Type, NodeType, CacheNodeType, Instance Type
Region	Yes	Region, Location
Version	No	Version, EngineVersion, Engine Version
Node count	No	NumNodes, NumCacheNodes, Node Count, Nodes
Cluster name	No	Cluster, Name, CacheClusterId

EKS

Field	Required	Recognized Column Names
Version	Yes	clusterVersion, KubernetesVersion, k8sVersion, Version
End of support date	Recommended	endOfSupportDate, end_of_support
End of extended support	No	endOfExtendedSupportDate, end_of_extended_support
Region	No	Region
Cluster name	No	clusterName, cluster_name, Cluster, Name

OpenSearch / Elasticsearch

Field	Required	Recognized Column Names
Instance type	Yes	Instance Type, InstanceType, Node Type
Region	Yes	Region, Location
Version	Recommended	Version, EngineVersion, Engine Version
Node count	No	NumNodes, Node Count, DataNodes, Nodes
Domain name	No	Domain, Name, DomainName, Cluster

Usage Examples

Generate reports from CSV inventories

"Calculate extended support costs for /path/to/rds_inventory.csv and save the report to /tmp/reports"

"Generate an EKS extended support report for /path/to/eks_clusters.csv and save to ~/Downloads"

"Process my ElastiCache inventory at /path/to/redis_clusters.csv and give me the cost breakdown"

"Calculate DocumentDB extended support costs for /path/to/docdb_instances.csv and save reports to ~/Desktop"

Quick single-instance cost checks

"What's the extended support cost for a db.r6g.xlarge PostgreSQL instance in us-east-1 with Multi-AZ?"

"How much would extended support cost for 10 EKS clusters over 6 months?"

"Calculate ElastiCache extended support for cache.r6g.large in eu-west-1 with 3 nodes"

"What's the OpenSearch extended support cost for an m7g.medium.search instance in us-west-2 with 5 nodes?"

Version lifecycle questions

"Which EKS versions are currently in extended support?"

"Show me all RDS and Aurora version extended support dates"

"When does PostgreSQL 14 enter extended support?"

"What OpenSearch versions have announced extended support dates?"

"Is Redis 6 eligible for ElastiCache extended support? When does it start?"

"List all services and versions that are entering extended support this year"

Data management

"Is the pricing data fresh?"

"Refresh the extended support pricing data"

"What services does the extended support calculator cover?"

Example Outputs

Single instance cost check

You: "What's the extended support cost for a db.r6g.xlarge PostgreSQL instance in us-east-1 with Multi-AZ?"

Kiro: The db.r6g.xlarge has 4 vCPUs. With Multi-AZ (2× multiplier), the effective vCPU count is 8.

	Per-vCPU Hourly Rate	Monthly	Yearly
Year 1	$0.10	$584	$7,008
Year 2	$0.10	$584	$7,008
Year 3	$0.20	$1,168	$14,016
Total 3-Year			$28,032

Version lifecycle query

You: "Which EKS versions are currently in extended support?"

Kiro: Here are the current EKS version support statuses:

Version	Standard Support End	Extended Support End	Status
EKS 1.30	Jul 22, 2025	Jul 22, 2026	In Extended Support
EKS 1.31	Nov 25, 2025	Nov 25, 2026	In Extended Support
EKS 1.32	Mar 22, 2026	Mar 22, 2027	In Extended Support
EKS 1.33	Jul 28, 2026	Jul 28, 2027	In Standard Support
EKS 1.34	Dec 1, 2026	Dec 1, 2027	In Standard Support
EKS 1.35	Mar 26, 2027	Mar 26, 2028	In Standard Support

Batch report

You: "Calculate extended support costs for /path/to/rds_inventory.csv and save the report to ~/Downloads"

Kiro: Processed 5,836 RDS/Aurora instances (130 skipped — unsupported engines).

2026 Extended Support Cost: $4,738,358 2027 Projected Cost: $17,260,574

Engine Version	Instances	Status	2026 Cost	2027 Projected
Aurora MySQL 5.7	130	In Extended Support	$1,596,755	$1,596,755
RDS MySQL 5.7	45	In Extended Support	$718,565	$791,343
RDS MySQL 8.0	828	Entering Aug 2026	$2,047,935	$4,915,043
RDS PostgreSQL 13	56	In Extended Support	$188,749	$226,499
RDS PostgreSQL 14	2,032	Entering Mar 2027	$0	$8,619,869
...

Reports saved:

• CSV: ~/Downloads/inventory_extended_support_20260427.csv
• HTML: ~/Downloads/inventory_extended_support_20260427.html

HTML report

The batch tools generate a styled HTML report that can be opened in any browser. It includes:

• Summary cards with current year and next year projected costs
• Category summary table (already in extended support / entering this year / entering next year)
• Version detail tables per category with instance counts, dates, and costs
• Region breakdown table

Tools

Single Instance Calculators

Quick cost estimates for individual resources.

Tool	Description
rds_cost	RDS/Aurora instance — per-vCPU-hour, Year 1/2/3 breakdown
elasticache_cost	ElastiCache Redis cluster — on-demand + premium tiers
eks_cost	EKS cluster(s) — extended vs standard support cost comparison
opensearch_cost	OpenSearch domain — NIH-based pricing
documentdb_cost	DocumentDB instance — per-vCPU-hour, Year 1/2/3 breakdown

If AWS doesn't publish a price for the requested instance type, pricing type, and region, the tool says no pricing was found rather than reporting $0.

Batch Report Generators

Process CSV inventories and generate executive-level CSV + HTML reports.

Tool	Description
rds_report	RDS/Aurora inventory → cost by engine/version, region, with current year + next year projections
elasticache_report	ElastiCache inventory → cost by version with premium tier projections
eks_report	EKS inventory → cost by version using end-of-support dates
opensearch_report	OpenSearch inventory → cost by engine version with NIH pricing
documentdb_report	DocumentDB inventory → cost with grace period handling
extended_support_report	Any mix of the inventories above → per-service reports plus a consolidated cross-service report

RDS and DocumentDB reports list any rows with no published price under no_pricing instead of counting them as $0.

All batch tools accept an output_dir parameter. When provided, they save:

• CSV report — summary, version breakdown, region breakdown, instance detail
• HTML report — styled executive report with summary cards, category tables, version detail

Reports categorize versions into:

• Already in Extended Support
• Entering Extended Support this year
• Entering Extended Support next year
• Future / Still in Standard Support

Utilities

Tool	Description
list_services	Show all supported services and pricing models
list_versions	List version lifecycle dates — shows which versions are in standard support, extended support, or future extended support with exact dates. Accepts a service name or "all"
list_billing_events	Informational list of AWS lifecycle events that change billing (from the Health API), across all services — including ones this calculator doesn't cost. Each event is flagged with whether the calculator can cost it
check_data	Check freshness of cached pricing and version data, and whether lifecycle dates are coming from the Health API or the built-in fallback
refresh_data	Force re-download all cached data

How Calculations Work

RDS / Aurora

1. Version eligibility: Fetched live from rds.describe_db_major_engine_versions() — returns exact standard support end, extended support start/end dates for each engine+version
2. Pricing: Fetched from the public AWS RDS Pricing API — extracts ExtendedSupport SKUs with rates by region, engine, and year. Provisioned instances use the per-vCPU-hour rate; Aurora Serverless v2 (db.serverless) uses the per-ACU-hour rate. The two are never substituted for each other
3. vCPU count: Looked up via ec2.describe_instance_types()
4. Cost formula: vCPUs × per-vCPU-hour rate × 730 hours/month × Multi-AZ multiplier (1 or 2)
5. Year determination: Month-by-month from the exact extended support start date. Year 1 = months 0-11, Year 2 = 12-23, Year 3 = 24+. Year 3 rates are typically 2× Year 1-2 for PostgreSQL/MySQL. For versions with more than three years of extended support, months after Year 3 bill at the Year 3 rate until extended support ends

ElastiCache

1. Version eligibility: Standard and extended support end dates per Redis OSS major version come from the AWS Health DescribeServiceLifecycle API; the Year 1/2/3 boundaries are derived from the standard support end date. If the API isn't available, built-in dates from the AWS docs are used
2. Pricing: On-demand price fetched from public ElastiCache Pricing API, then premium applied
3. Premium tiers: Year 1-2: 80% of on-demand price. Year 3: 160% of on-demand price
4. Cost formula: on-demand hourly × premium % × num_nodes × 730 hours/month

EKS

1. Version eligibility: Fetched live from eks.describe_cluster_versions() — returns endOfStandardSupportDate, endOfExtendedSupportDate, and status per version
2. Pricing: Additional hourly rate per region fetched from public AWS EKS Pricing API — no credentials needed. Currently $0.50/cluster/hour additional across all regions
3. Cost formula: extended_support_rate × 730 hours/month (only for months after end-of-support date)
4. CSV dates: The batch tool reads endOfSupportDate and endOfExtendedSupportDate columns from the CSV (if present) to determine per-cluster support timelines. These columns are typically populated by data collection tools like Dante/TSC

OpenSearch / Elasticsearch

1. Version eligibility: Standard and extended support end dates per Elasticsearch/OpenSearch version come from the AWS Health DescribeServiceLifecycle API. If the API isn't available, built-in dates from the AWS blog post are used
2. Pricing: NIH rate per region fetched from public AWS OpenSearch Pricing API — no credentials needed
3. Cost formula: NIH_rate × normalization_factor × num_nodes × 730 hours/month
4. Normalization factors: nano=0.25, micro=0.5, small=1, medium=2, large=4, xlarge=8, 2xlarge=16, etc. Source: AWS blog post

DocumentDB

1. Version eligibility: Only v3.6 has extended support. Its standard and extended support end dates come from the AWS Health DescribeServiceLifecycle API, with built-in dates from the AWS docs as a fallback
2. Pricing: Per-vCPU-hour rates fetched from the public DocumentDB Pricing API, matched by instance family
3. Grace period: Standard support ends Mar 30, 2026 but extended support billing starts Jul 1, 2026 (3-month grace). The API doesn't publish this, so it's built in
4. Cost formula: vCPUs × per-vCPU-hour rate × 730 hours/month × Multi-AZ multiplier
5. Rate bands: Years 1 and 2 share one rate; Year 3 has a higher rate. Both come from the pricing API

Data Freshness

All fetched data is cached locally for 24 hours in a cache/ directory next to the server script.

Data	Source	Requires Credentials	Cache TTL
RDS extended support pricing	AWS Pricing API (public)	No	24 hours
ElastiCache on-demand pricing	AWS Pricing API (public)	No	24 hours
DocumentDB extended support pricing	AWS Pricing API (public)	No	24 hours
OpenSearch NIH rates	AWS Pricing API (public)	No	24 hours
EKS extended support rate	AWS Pricing API (public)	No	24 hours
EC2 vCPU counts	ec2.describe_instance_types()	Yes	24 hours
RDS version lifecycle	rds.describe_db_major_engine_versions()	Yes	24 hours
EKS version lifecycle	eks.describe_cluster_versions()	Yes	24 hours
ElastiCache, OpenSearch, DocumentDB lifecycle dates	health.describe_service_lifecycle() (AWS Health)	Yes	24 hours

If the Health API can't be reached, ElastiCache, OpenSearch, and DocumentDB fall back to built-in dates.

Use check_data to see cache status and which lifecycle source is in use. Use refresh_data to force re-download.

First tool call on a fresh install triggers automatic data fetch — no manual setup needed.