---
title: "AWS Redshift MCP: review, setup & limitations · Undominated.ai"
canonical: https://undominated.ai/mcp-servers/awslabs-redshift/
description: "Discover Redshift schemas and run bounded SQL with read-only execution as the default."
---

# AWS Redshift MCP: review, setup & limitations · Undominated.ai

> Discover Redshift schemas and run bounded SQL with read-only execution as the default.

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

AWS DATA SERVICES / AWS Labs

# AWS Redshift MCP

Discover Redshift schemas and run bounded SQL with read-only execution as the default.

 [See setup guidance ↓](#setup)[Original source ↗](https://github.com/awslabs/mcp)

SOURCE REVIEW

 Reviewed 2026-10-14
 Evidence 5 linked sources
 Publisher AWS Labs
 Licence [Apache-2.0 ↗](https://github.com/awslabs/mcp/blob/8840003820a78dd018561d55c7ab12fcdd708464/LICENSE)
 Revision 8840003820a7
 [Read what was—and wasn’t—checked ↓](#review)

“MCP server for Amazon Redshift metadata discovery and SQL execution.”

 [AWS Labs · upstream description ↗](https://github.com/awslabs/mcp/blob/8840003820a78dd018561d55c7ab12fcdd708464/src/redshift-mcp-server/README.md) Our analysis follows below.

01 / THE REASONING

## Why this made the selection.

 - Specialized publisher integration with documented IAM and data access boundaries.
- Combines provisioned/serverless discovery with transaction-backed read-only SQL and bounded result handling.
- The README discovers Redshift clusters and workgroups and runs SQL, with read-only as the startup default.
- It says the statement guard is not a security boundary and that database privileges are the durable control.

### A good fit for

 - Investigate warehouse schema and aggregate data using a dedicated read-only database role.
- Review cluster diagnostics and query plans before proposing a separately approved data change.
- Exploring Redshift metadata and running read-only SQL
- A write workload only when ACCESS_MODE is set to read-write before the process starts

### Weigh up before choosing

 - IAM permission to use the Data API does not grant database SELECT privileges; provision both and avoid a superuser identity.
- MAX_RESULT_ROWS bounds rows, not bytes. Oversized results are refused, while wide rows can still produce large responses.
- Read-write mode autocommits statements outside named transactions. Confirmation requires a capable MCP client; UNSAFE_SKIP_WRITE_CONFIRMATION removes that control.
- DEBUG logs include SQL and engine errors can contain sensitive values. Independent calls do not retain session-local temporary tables or settings.
- The monorepo now recommends Agent Toolkit for AWS for production. This specialized local server remains documented; its leaf and root documentation were checked for deprecation.
- Read-write cannot be enabled per call. A typo falls back to read-only.
- The SQL guard does not inspect intent. A single DROP, DELETE, or UPDATE without WHERE can still run once confirmed.
- Clients that cannot show the write confirmation are refused rather than allowed to write unconfirmed, unless the unsafe skip flag is set.
- A pin does not verify transitive dependencies. Documented uvx examples use a floating package tag, not the reviewed commit.
- This review did not execute the resource.
- The server was not started and its tools were not called.

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/redshift-mcp-server/README.md
- src/redshift-mcp-server/pyproject.toml
- src/redshift-mcp-server/awslabs/redshift_mcp_server/server.py
- Root README.md and LICENSE
- LICENSE

### Our findings

 - The pinned entrypoint resolves access mode once at startup and refuses oversized results rather than silently truncating.
- Source and README describe read-only transaction wrapping, explicit write mode and confirmation failure when the client cannot prompt. These controls were inspected, not bypass-tested.
- The server discovers provisioned clusters and serverless workgroups and executes SQL. It is read-only unless ACCESS_MODE=read-write is selected at startup.
- The mode cannot change for the life of the process. redshift-serverless:GetWorkgroup and redshift:GetClusterCredentials are described as not needed.
- Single-statement checks do not make a write safe. The README says to treat that guard as defense in depth, not a security boundary, and to rely on the database user's privileges.

### Not established by this review

 - Server startup, authentication, live tool execution and security isolation were not tested.
- Published package/container parity with a source revision and hosted implementation internals were not tested.
- The server was not started and its tools were not called.
- A pin does not verify transitive dependencies.
- No tools/list or tool call was made.

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/8840003820a78dd018561d55c7ab12fcdd708464/src/redshift-mcp-server/README.md)

DOCUMENTED COMMAND

 uvx awslabs.redshift-mcp-server@latest Copy command ↗

Copying does not execute this command. It may retrieve a newer version than the reviewed source.

 - Configure a dedicated AWS profile and read-only database grants for the intended cluster or serverless workgroup.
- Grant the documented Redshift discovery and Data API actions, including BatchExecuteStatement for the current transaction-backed path.
- Launch in read-only mode, discover database/schema/table names, then use explicit columns, predicates and result limits.

### Before you start

 - Python 3.10 or newer, uv/uvx and an MCP client that can launch a local stdio process.
- An AWS account, configured credentials and access to the intended region and service resources.

### Documented configuration example

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

 {
 "mcpServers": {
 "awslabs-redshift": {
 "command": "uvx",
 "args": [
 "awslabs.redshift-mcp-server@latest"
 ],
 "env": {
 "AWS_PROFILE": "redshift-reader",
 "AWS_DEFAULT_REGION": "us-east-1",
 "ACCESS_MODE": "read-only",
 "UNSAFE_SKIP_WRITE_CONFIRMATION": "false",
 "MAX_RESULT_ROWS": "1000",
 "LOG_LEVEL": "WARNING"
 }
 }
 }
} Copy configuration ↗

### Compatibility

Python 3.10 or newer, uv/uvx and an MCP client that can launch a local stdio process. · An AWS account, configured credentials and access to the intended region and service resources.

### Implementation

Python

### Transports

stdio

### AWS data services

 - Discovers provisioned clusters and serverless workgroups; executes SQL as the database user associated with the AWS identity.
- Cluster diagnostic review additionally needs sys:monitor or a superuser; use the narrower monitoring role.

### Cost model

AWS service, model, query and storage charges apply; running the MCP process does not include those services.

04 / FOLLOW THE EVIDENCE

## The source trail.

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

 - [Publisher setup, capabilities and operating boundaries ↗](https://github.com/awslabs/mcp/blob/8840003820a78dd018561d55c7ab12fcdd708464/src/redshift-mcp-server/README.md) Checked 2026-10-14 https://github.com/awslabs/mcp/blob/8840003820a78dd018561d55c7ab12fcdd708464/src/redshift-mcp-server/README.md Supports: summary, install, configExample, access, limitations, bestFor, review, compatibility, transports
- [Pinned package and implementation ↗](https://github.com/awslabs/mcp/blob/8840003820a78dd018561d55c7ab12fcdd708464/src/redshift-mcp-server/pyproject.toml) Checked 2026-10-14 https://github.com/awslabs/mcp/blob/8840003820a78dd018561d55c7ab12fcdd708464/src/redshift-mcp-server/pyproject.toml Supports: summary, install, configExample, access, limitations, bestFor, review, compatibility, transports
- [Pinned package and implementation ↗](https://github.com/awslabs/mcp/blob/8840003820a78dd018561d55c7ab12fcdd708464/src/redshift-mcp-server/awslabs/redshift_mcp_server/server.py) Checked 2026-10-14 https://github.com/awslabs/mcp/blob/8840003820a78dd018561d55c7ab12fcdd708464/src/redshift-mcp-server/awslabs/redshift_mcp_server/server.py Supports: summary, install, configExample, access, limitations, bestFor, review, compatibility, transports
- [Pinned package and implementation ↗](https://github.com/awslabs/mcp/blob/8840003820a78dd018561d55c7ab12fcdd708464/README.md) Checked 2026-10-14 https://github.com/awslabs/mcp/blob/8840003820a78dd018561d55c7ab12fcdd708464/README.md Supports: summary, install, configExample, access, limitations, bestFor, review, compatibility, transports
- [Applicable licence ↗](https://github.com/awslabs/mcp/blob/8840003820a78dd018561d55c7ab12fcdd708464/LICENSE) Checked 2026-10-14 https://github.com/awslabs/mcp/blob/8840003820a78dd018561d55c7ab12fcdd708464/LICENSE Supports: license

KEEP COMPARING

## Other approaches to consider.

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

 [### AWS Keyspaces MCP ↗ Inspect Cassandra-compatible schemas and run SELECT queries with dedicated database credentials.](/mcp-servers/awslabs-keyspaces/)[### AWS Labs Amazon Kendra Index ↗ Queries an existing Kendra index and returns document excerpts, source URIs and confidence labels for retrieval workflows.](/mcp-servers/awslabs-kendra-index/)[### AWS Labs Amazon MQ ↗ Inspects and manages Amazon MQ brokers, with additional RabbitMQ management operations.](/mcp-servers/awslabs-amazon-mq/)

 [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/)
