AWS Billing

What AWS Billing MCP server does

The AWS Billing MCP server gives AI agents access to selected AWS billing data through a hosted MCP service. It connects to AWS using IAM credentials linked by the user and returns normalized results from AWS Cost Explorer and AWS Cost Anomaly Detection.

The service is read-only. Agents can inspect spending, compare costs by service, request a forecast, and review detected anomalies, but no tool changes AWS billing records or other source data. This fits assistants that need to answer questions such as which services drove spending over a recent period or what AWS spend may look like next month.

How it works

Clients connect to the hosted MCP over HTTP at https://api.mcp.ai/p_aws_billing. After the user links AWS credentials, the available tools query billing data on the user’s behalf. Authentication uses a passwordless magic-link flow rather than a password.

The server returns normalized tool results for cost and usage, forecasts, and anomalies. The selected LLM host receives the data returned by the tools, so the host is part of the processing path. The repository identifies the service as hosted and proprietary, while the public wrapper, manifests, documentation, and skills are released under MIT.

Setup and configuration

To install the AWS Billing MCP server, add the hosted HTTP endpoint to an MCP client. Claude Web and Claude Desktop can use the custom connector flow. Cursor and VS Code have installation links in the project documentation. Other clients can connect if they support MCP over HTTP.

No server-side environment variables are documented as required. Configuration consists of the MCP endpoint and the user’s linked AWS credentials. The README provides a client-specific installation guide in INSTALL.md, and the endpoint remains the same for supported HTTP clients.

Tools and capabilities

Four tools are available:

• aws_billing_list_accounts lists billing credential records associated with the installation.
• aws_billing_cost returns normalized Cost Explorer cost and usage data, including a daily breakdown and leading services.
• aws_billing_forecast returns a normalized Cost Explorer forecast for the following 30 days or for a supplied date range.
• aws_billing_anomalies returns normalized Cost Anomaly Detection anomalies for a specified date range.

Together, these tools support recent cost review, service-level spending analysis, forward-looking estimates, and anomaly checks.

Limitations and notes

The AWS Billing MCP server only reads data; it does not provide mutation tools. Access depends on the IAM credentials connected by the user and on the selected client’s MCP over HTTP support.

Data returned by the tools is sent to the LLM host chosen by the user. That host is identified as a subprocesser outside the service’s control, so users should review the host’s data and training settings. The hosted server itself is proprietary even though the surrounding public repository is MIT-licensed.