AWS Tool Turns 1,122 Open Datasets Into AI Queries

AWS has released an open-source AWS’s open source RODA MCP server announcement, and the tool puts the company’s Registry of Open Data behind a single chat prompt in Kiro, Claude Code, and any other AI assistant that speaks Anthropic’s Model Context Protocol.

The registry just crossed 1,122 entries in a public catalog that includes data from NASA, NOAA, the National Institutes of Health, the Allen Institute, and “hundreds of others” committed to open access. Hosting the catalog on AWS lets researchers analyze cloud-stored files without paying for their own storage; they only pay for the compute they use, and the storage cost is sponsored. The new RODA MCP server adds a conversational front door to that catalog. Find a dataset, ask for its license and metadata, and pull a sample file, all from inside an AI assistant.

Three Steps from a Question to a Dataset

The RODA server exposes tools for three jobs: discovery, exploration, and evaluation. Ask “search for datasets to study ocean temperature” and the assistant returns matching entries with their descriptions and licensing terms. Ask “get more details about 1000 Genomes data” and the assistant pulls full metadata plus related resources from the same catalog.

Once a researcher has a candidate, the evaluation step is where time used to vanish. The new tool can list the underlying S3 bucket structure, name the file formats, and sample a small read of a chosen file so a scientist can decide whether to commit to deeper work. AWS describes the framing directly in the same announcement:

turning the lengthy multi-step discovery process into a single conversation.

In practice that means a researcher can move from “I want NOAA land surface temperature data under a Creative Commons license” to “preview the file structure of the 1000 Genomes dataset and sample a README” without leaving the chat box.

What a Catalog of More Than 1,100 Datasets Looks Like

The catalog is the spine of the launch. AWS tracks the steady climb with each new dataset wave. The flagship figures:

• 1,122 datasets are now freely available to anyone.
• Over 400 Petabytes of data sit in the catalog, hosted on AWS and ready to query.
• Data comes from organizations like NOAA, NASA, the Allen Institute, NIH, and Biohub, with hundreds of others committed to open access.
• The catalog spans genomics, satellite imagery, climate and weather, natural language processing, autonomous vehicles, and more.

The growth is no recent blip. A separate milestone post on the open data registry documents that it “more than doubled in size, growing from 556 to over 1,100 datasets, representing a 102% increase in datasets.” Each new wave brings in more domains, more providers, and more petabytes.

Anthropic’s Protocol, AWS’s Open-Source Server

The RODA server ships as an Apache 2.0-licensed project on a public GitHub repository. It works because it speaks an existing protocol rather than inventing one. Anthropic’s Model Context Protocol is the standard the server targets, and that choice lets a single release plug into the growing universe of MCP-compatible AI tools without bespoke integrations per assistant.

The roster of clients already speaks MCP:

• Kiro
• Claude Code
• Cline
• Cursor
• Windsurf
• Claude Desktop

Anthropic introduced MCP as an open standard for connecting AI assistants to data sources and tools, and the AWS Labs open-source MCP directory confirms how widely AWS has committed to the protocol. The directory lists 62 open-source MCP servers in total, and the RODA server appears in the data-and-analytics section of that catalog. AWS Labs hosts the broader suite.

Three Queries That Already Work

The launch post shows three working prompts that any configured assistant can answer against the registry:

1. “What genomics datasets are available under a Creative Commons license?”
2. “Show me datasets related to land surface temperature.”
3. “Preview the file structure of the 1000 Genomes dataset and sample a README.”

Getting the server into an assistant takes one command. Install it via uv with uvx awslabs.roda-mcp-server@latest. Configure Kiro at ~/.kiro/settings/mcp.json. Add it to Claude Code with claude mcp add roda-mcp uvx awslabs.roda-mcp-server@latest. The full snippets and prerequisites sit on the AWS Open Source blog.

How It Plugs Into SageMaker and Bedrock

The point of hosting these datasets on AWS rather than on a generic S3 store is what comes next. AWS lists the tools it expects researchers to combine with the catalog: “Amazon EC2, Amazon SageMaker, Amazon Bedrock, Amazon Athena, AWS Lambda, and Amazon EMR.”

Each name is a different way to spend compute on the same data. A researcher can train a model in SageMaker, query a table with Athena, run a serverless workflow in Lambda, or invoke a foundation model in Bedrock, all without paying the catalog for a local copy of the dataset. AWS prices the round honestly: storage is sponsored under the Open Data Sponsorship Program, and the user “only pay[s] for the compute they use.” A vector search, a Bedrock agent, and a SageMaker training run all draw on the same source without renegotiating it. The RODA MCP server is the conversational layer that lets a researcher start that workflow without first hand-finding files.

Open Data Now Sits Inside the AI Stack

The Registry of Open Data still exists as a browsable catalog at registry.opendata.aws. The page lists 1,148 matching datasets visible to a visitor today. The MCP server is a thin wrapper around that catalog, not a replacement for it; the chat interface is now one of two ways in.

The launch also slots AWS into a wider race to own the data layer that AI agents rely on. Salesforce, Microsoft, SAP, and Snowflake are all pushing into the same space, as covered in a related piece on the race for AI’s trusted data layer. AWS is taking the open-protocol route. Instead of locking the catalog behind a proprietary search box, the company publishes the server under Apache 2.0 and reuses Anthropic’s MCP as a lingua franca for AI assistants. The source-side description still reads “the Registry of Open Data makes it easy to find datasets made publicly available through AWS services.” The MCP server is now the conversational entry point to that promise, available to anyone who can install a Python package.