---
title: "Amazon Bedrock Knowledge Base Retrieval MCP Server: review, setup & limitations · Undominated.ai"
canonical: https://undominated.ai/mcp-servers/awslabs-bedrock-kb-retrieval/
description: "Lists and queries Amazon Bedrock Knowledge Bases that are tagged for MCP use, with data-source filtering, optional reranking and agentic retrieval on managed knowledge bases."
---

# Amazon Bedrock Knowledge Base Retrieval MCP Server: review, setup & limitations · Undominated.ai

> Lists and queries Amazon Bedrock Knowledge Bases that are tagged for MCP use, with data-source filtering, optional reranking and agentic retrieval on managed knowledge bases.

[← Explore all mcp servers](/mcp-servers/)

RETRIEVAL AND GROUNDING / Amazon Web Services

# Amazon Bedrock Knowledge Base Retrieval MCP Server

Lists and queries Amazon Bedrock Knowledge Bases that are tagged for MCP use, with data-source filtering, optional reranking and agentic retrieval on managed knowledge bases.

 [See setup guidance ↓](#setup)[Original source ↗](https://github.com/awslabs/mcp/tree/e1abc4b3b16a6da4c86ccbfc89de7bd864c5679e/src/bedrock-kb-retrieval-mcp-server)

SOURCE REVIEW

 Reviewed 2026-10-13
 Evidence 4 linked sources
 Publisher Amazon Web Services
 Licence [Apache-2.0 ↗](https://github.com/awslabs/mcp/blob/e1abc4b3b16a6da4c86ccbfc89de7bd864c5679e/src/bedrock-kb-retrieval-mcp-server/LICENSE)
 Revision e1abc4b3b16a
 [Read what was—and wasn’t—checked ↓](#review)

“MCP server for accessing Amazon Bedrock Knowledge Bases”

 [Amazon Web Services · upstream description ↗](https://github.com/awslabs/mcp/blob/e1abc4b3b16a6da4c86ccbfc89de7bd864c5679e/src/bedrock-kb-retrieval-mcp-server/README.md) Our analysis follows below.

01 / THE REASONING

## Why this made the selection.

 - Published by AWS in awslabs/mcp; the README at this commit exposes only knowledge bases carrying the tag key mcp-multirag-kb with value true, so each knowledge base is opted in by the account owner.
- The README states reranking availability differs per model and region (amazon.rerank-v1:0 is not offered in us-east-1, cohere.rerank-v3-5:0 is) and that the server validates the region and model pair before calling.

### A good fit for

 - Answering questions from documents already indexed in Amazon Bedrock Knowledge Bases, with citations
- Retrieving permission-filtered content from ACL-aware sources such as SharePoint, OneDrive or Confluence by passing a user_id

### Weigh up before choosing

 - AgenticQueryKnowledgeBases invokes a foundation model to write a cited answer unless generate_response=false; the README says it costs materially more per call than QueryKnowledgeBases and works on managed knowledge bases only.
- Results with IMAGE content type are not included, and content from ACL-aware data sources is inaccessible unless a user_id is passed.
- The README configures awslabs.bedrock-kb-retrieval-mcp-server@latest, which is not pinned to this commit. No knowledge base was queried.

02 / THE REVIEW RECORD

## What we actually inspected.

Source review has boundaries. A clear record is more useful than a “safe” badge.

### Material inspected

 - src/bedrock-kb-retrieval-mcp-server/README.md at e1abc4b3b16a6da4c86ccbfc89de7bd864c5679e
- src/bedrock-kb-retrieval-mcp-server/pyproject.toml at the same commit
- src/bedrock-kb-retrieval-mcp-server/LICENSE (byte-identical to the repository root LICENSE) and src/bedrock-kb-retrieval-mcp-server/NOTICE at the same commit

### Our findings

 - NOTICE names awslabs.bedrock-kb-retrieval-mcp-server, Copyright Amazon.com, Inc. or its affiliates.
- Reranking is off by default (BEDROCK_KB_RERANKING_ENABLED=false) and needs bedrock:Rerank and bedrock:InvokeModel on both the caller and the Knowledge Bases service role.
- pyproject.toml names awslabs.bedrock-kb-retrieval-mcp-server 1.1.2, Python >=3.10, license Apache-2.0.

### Not established by this review

 - uvx launch
- Any Retrieve, Rerank or AgenticRetrieveStream call
- ACL-aware retrieval with user_id

The review applies to the material and revision named here. A newer upstream release can change its behavior.

03 / PUT IT TO WORK

## Connect a server deliberately.

[Upstream setup instructions ↗](https://github.com/awslabs/mcp/blob/e1abc4b3b16a6da4c86ccbfc89de7bd864c5679e/src/bedrock-kb-retrieval-mcp-server/README.md)
 - The README gives a client configuration block, not a one-line command: command uvx with args awslabs.bedrock-kb-retrieval-mcp-server@latest. It was not run.
- Set AWS_PROFILE and AWS_REGION, tag each knowledge base to expose with mcp-multirag-kb=true (or set KB_INCLUSION_TAG_KEY), and grant the IAM permissions the README lists.

### Before you start

 - uv and Python 3.10
- AWS credentials with access to Amazon Bedrock Knowledge Bases
- At least one knowledge base tagged for MCP use

### Documented configuration example

Use the client and authentication context described in the setup guidance. Configuration formats vary between hosts.

 {
 "mcpServers": {
 "awslabs.bedrock-kb-retrieval-mcp-server": {
 "command": "uvx",
 "args": [
 "awslabs.bedrock-kb-retrieval-mcp-server@latest"
 ],
 "env": {
 "AWS_PROFILE": "your-profile-name",
 "AWS_REGION": "us-east-1",
 "FASTMCP_LOG_LEVEL": "ERROR",
 "KB_INCLUSION_TAG_KEY": "optional-tag-key-to-filter-kbs",
 "BEDROCK_KB_RERANKING_ENABLED": "false"
 },
 "disabled": false,
 "autoApprove": []
 }
 }
} Copy configuration ↗

### Compatibility

Kiro · Cursor · VS Code

### Implementation

Python

### Transports

stdio

### Local process, AWS credentials

 - Lists knowledge bases and data sources and runs retrieval queries with the selected AWS profile
- Optional reranking and agentic retrieval invoke Bedrock models billed to the account

### Cost model

Apache-2.0 server. Amazon Bedrock retrieval, reranking and model invocation are billed by AWS; no rate was reviewed.

04 / FOLLOW THE EVIDENCE

## The source trail.

Our notes are separate from the original resource. Check upstream before adopting a new version.

 - [Bedrock KB Retrieval MCP README ↗](https://github.com/awslabs/mcp/blob/e1abc4b3b16a6da4c86ccbfc89de7bd864c5679e/src/bedrock-kb-retrieval-mcp-server/README.md) Checked 2026-10-13 https://github.com/awslabs/mcp/blob/e1abc4b3b16a6da4c86ccbfc89de7bd864c5679e/src/bedrock-kb-retrieval-mcp-server/README.md Supports: Tag-based opt-in, Reranking and agentic retrieval requirements, Client configuration, Limitations
- [Bedrock KB Retrieval MCP LICENSE ↗](https://github.com/awslabs/mcp/blob/e1abc4b3b16a6da4c86ccbfc89de7bd864c5679e/src/bedrock-kb-retrieval-mcp-server/LICENSE) Checked 2026-10-13 https://github.com/awslabs/mcp/blob/e1abc4b3b16a6da4c86ccbfc89de7bd864c5679e/src/bedrock-kb-retrieval-mcp-server/LICENSE Supports: Apache License 2.0
- [Bedrock KB Retrieval MCP NOTICE ↗](https://github.com/awslabs/mcp/blob/e1abc4b3b16a6da4c86ccbfc89de7bd864c5679e/src/bedrock-kb-retrieval-mcp-server/NOTICE) Checked 2026-10-13 https://github.com/awslabs/mcp/blob/e1abc4b3b16a6da4c86ccbfc89de7bd864c5679e/src/bedrock-kb-retrieval-mcp-server/NOTICE Supports: Copyright Amazon.com, Inc. or its affiliates
- [Bedrock KB Retrieval MCP pyproject.toml ↗](https://github.com/awslabs/mcp/blob/e1abc4b3b16a6da4c86ccbfc89de7bd864c5679e/src/bedrock-kb-retrieval-mcp-server/pyproject.toml) Checked 2026-10-13 https://github.com/awslabs/mcp/blob/e1abc4b3b16a6da4c86ccbfc89de7bd864c5679e/src/bedrock-kb-retrieval-mcp-server/pyproject.toml Supports: Package version 1.1.2, Python >=3.10

KEEP COMPARING

## Other approaches to consider.

Related by category or shared topics. These are alternatives to inspect, not a measured quality order.

 [### Pinecone Assistant MCP Server ↗ Retrieves document snippets from a named Pinecone Assistant for an MCP client through one tool, run as a locally built Docker image or Rust binary.](/mcp-servers/pinecone-assistant-mcp/)[### Vectara MCP Server ↗ Runs Vectara RAG queries and semantic search over named corpora, and calls Vectara's hallucination-correction and factual-consistency APIs, from an MCP client.](/mcp-servers/vectara-mcp/)[### AWS Bedrock Custom Model Import MCP Server ↗ Creates and tracks Amazon Bedrock custom model import jobs, and lists, inspects or deletes imported models, from an MCP client. Read-only unless started with --allow-write.](/mcp-servers/awslabs-bedrock-custom-model-import/)

 [AI Tools ↗](/tools/)[Skills ↗](/skills/)[Agents ↗](/agents/)[MCP Servers ↗](/mcp-servers/)[Workflows ↗](/workflows/)

## Continue your investigation

 - [Inspect resource evidence](/tools/)
- [Choose a host](/agents/)
- [Inspect reusable workflows](/skills/)
- [Use Undominated data](/api/)
