read_sections returns the section you named, search_table returns the rows that matched inside a quota or pricing table, and read_documentation pages through anything long. That matters more here than for most documentation servers, because AWS reference pages routinely carry tables longer than the prose around them.
A documentation server for AWS docs. It searches through AWS's own documentation search API and converts pages to markdown, with tools for taking only the part of a page you need.
- AWS documentation searched by phrase, narrowed by guide type, product type and search intent — search_documentation
- Any documentation page fetched and converted to markdown, resumable through start_index for pages too long to return at once — read_documentation
- Named sections extracted from a page rather than the whole document, matched without regard to case or whitespace — read_sections
- Matching rows pulled out of a large table — service quotas, pricing, supported-model lists — as JSON instead of the entire table — search_table
- Related pages suggested for a URL you are already reading — recommend
It reads public AWS documentation, so there are no credentials to set up: install uv, install Python with uv python install 3.10, and run awslabs.aws-documentation-mcp-server through uvx. AWS_DOCUMENTATION_PARTITION selects the partition — set it to aws-cn for AWS China documentation, where search, table search and recommendations are not offered. Behind a corporate proxy that filters on User-Agent, MCP_USER_AGENT replaces the default one.
One command — uvx awslabs.aws-documentation-mcp-server