Amazon Kendra Index MCP Server

Overview

An AWS Labs Model Context Protocol (MCP) server that connects to Amazon Kendra indexes so MCP-enabled assistants can use Kendra as an additional context source for retrieval-augmented generation (RAG) and enterprise search.

• Category: Search & Discovery MCP Servers
• Vendor / Brand: Amazon Web Services
• Primary use cases:
  • Augment conversational and coding assistants with enterprise search results from Amazon Kendra indices
  • Query existing Kendra indexes for documentation, knowledge base, and RAG workflows

Features

Core Capabilities

• Connects MCP-enabled chatbots to Amazon Kendra indexes as RAG data sources.
• Enhances responses from coding and AI assistants (e.g., Cline, Cursor, Windsurf, Amazon Q Developer) by injecting Kendra search results into context.
• Supports querying either a default Kendra index or a specific index referenced in the user prompt.
• Can list available Amazon Kendra indexes in your AWS account to help select an appropriate index at runtime.

Tools

1. KendraQueryTool

• Function: Uses a natural-language query to search an Amazon Kendra index and retrieve additional context for the assistant’s response.
• Behavior:
  • Queries a default index, or
  • Queries a specific index if the user names it in the prompt.
• Required parameters:
  • query (str) – The user’s search query.
• Optional parameters:
  • indexId (str) – ID of the Kendra index to query.
  • region (str) – AWS region of the target Kendra index.
• Example usage prompts:
  • “Can you help me understand how to implement a progress event in the CreateHandler using Java? Use the KendraQueryTool to gain additional context.”
  • “Can you use the test-kendra-index to help answer the following questions…?”

2. KendraListIndexesTool

• Function: Lists Amazon Kendra indexes in the AWS account.
• Behavior:
  • By default, lists indices in regions defined via environment variables in the MCP configuration.
  • Region can be overridden in the user prompt.
• Optional parameters:
  • region (str) – AWS region whose Kendra indexes should be listed.
• Example usage prompt:
  • “Can you list the Kendra Indexes in my account in the us-west-2 region?”

Requirements

Pre-Requisites

1. An AWS account.
2. An Amazon Kendra index containing the documentation or data you want to use for RAG.
3. uv installed (from Astral) for environment and dependency management.
4. Python 3.10 installed (e.g., via uv python install 3.10).

IAM & AWS Configuration

1. Create or use an IAM user in your AWS account.
2. Attach a policy with, at minimum:
   • kendra:Query
   • kendra:ListIndices
   • (Alternatively, the managed policy AmazonKendraFullAccess, following least-privilege principles.)
3. Configure AWS credentials (access key ID and secret access key) in the environment using aws configure.

Editor / Client Integration

• Installable as an MCP server in compatible clients such as Cursor and VS Code via provided install links/config.

Pricing

No specific pricing information is provided for the MCP server itself. Usage will typically incur standard AWS charges for:

• Amazon Kendra (index storage, queries, and related operations), and
• Any other underlying AWS resources used by your account.

Consult AWS pricing pages for Amazon Kendra and related services for detailed cost information.