How to use Claude with Redshift

Redshift is the warehouse most teams on AWS end up with. It's reliable, it scales, and it integrates well with the rest of the AWS ecosystem. But when it comes to connecting AI agents, Redshift is underserved compared to Snowflake or BigQuery. Snowflake has Cortex Analyst. BigQuery has Conversational Analytics. Redshift has... a collection of community MCP servers and a fragmented set of AWS AI services. Here's how to actually get your team querying Redshift through Claude.

Option 1: Redshift MCP server (direct connection)

AWS released an official Redshift MCP server that connects Claude to your cluster. It supports cluster discovery, metadata exploration, and read-only query execution. There are also several community-maintained alternatives on GitHub.

// claude_desktop_config.json
{
  "mcpServers": {
    "redshift": {
      "command": "uvx",
      "args": ["awslabs.redshift-mcp-server@latest"],
      "env": {
        "AWS_PROFILE": "your-profile",
        "AWS_REGION": "us-east-1"
      }
    }
  }
}

This gives Claude tools to list clusters, explore schemas, and run SQL. It works for a data engineer exploring a familiar schema. It does not work for a team.

Limitations

• No schema context: Claude sees table and column names but doesn't know what they mean. If your table is stg_events_v2, Claude has to guess the schema.
• No shared definitions: "Active user" or "MRR" mean different things to different people. Claude interprets them based on phrasing.
• Per-person setup: Everyone configures their own connection and manages their own AWS credentials.
• No access controls: Everyone with the connection string can query everything. No way to restrict tables by team or role.
• No audit trail: No visibility into what questions your team is asking or what queries are running.

Option 2: AWS native AI (Amazon Q, Bedrock)

AWS has several AI services that can work with Redshift: Amazon Q for business intelligence, Bedrock for LLM access, and Redshift ML for in-warehouse predictions. But there's no single "ask your Redshift data questions in plain English" product like Snowflake Cortex Analyst.

You can stitch together a solution using Bedrock + Lambda + Redshift Data API, but that's a custom engineering project. You're building and maintaining the context layer, the access controls, and the query pipeline yourself.

Option 3: Redshift + Claude via Contextflo

Contextflo connects to your Redshift cluster using a read-only database user. It reads your schema, source code, and docs, auto-generates context for every table and column, and serves that context to Claude through MCP.

Your team asks questions in Claude. Claude uses the context layer to write accurate Redshift SQL. Contextflo runs the query and returns results. Everyone gets the same definitions, the same access controls, and everything is logged.

It's free for one user and one data source, so you can validate the whole flow on your own cluster before rolling it out to the team.

Comparison

Which to choose

The Redshift MCP server is fine for a data engineer who wants to query Redshift from Claude without switching tools. It's fast to set up and free.

AWS native makes sense if you have a dedicated data engineering team and want everything within the AWS ecosystem, but expect a multi-week build.

Contextflo is for teams where multiple people need to ask questions and get consistent, controlled answers - without a custom engineering project.

Frequently asked questions

Free for one user and one data source. Setup takes about 15 minutes.