Congressional Tech · Use with AI

Congress Explorer MCP

Congressional data.
Your questions. Your assistant.

Find hearings, witnesses, bills, members, recordings, and related documents through your AI assistant.

Free & open source · macOS and Linux · No data API key

Quick setup

One server.
Your choice of assistant.

Every setup option installs the same versioned PyPI package. The first question downloads the published metadata snapshot. Later sessions reuse it. Queries run on your computer.

About 65 MiB of data as of October 2026, plus Python and dependencies. Windows is not yet supported.

Other clients & manual setup ↓

Claude Desktop chat

Download the extension, open it with Claude Desktop, and click Install. Use a current version with support for uv-based extensions; the app manages Python and dependencies.

Download .mcpb

If opening the file does not show an installer, go to Settings → Desktop app → Extensions → Advanced settings → Install Extension.

Claude Code & Cowork

Install uv on the computer running the plugin, then use this command with a current Claude Code:

claude plugin install congress-explorer --marketplace civictechdc/congressional-tech

In Cowork desktop, open Customize → Plugins and upload the plugin ZIP. Start a local task, then run /congress-explorer:explore-congress.

Download plugin ZIP

The plugin's local server runs in Claude Code and Cowork desktop. For Claude Desktop chat, use the extension above.

ChatGPT & Codex desktop

Install uv and the Codex CLI, then add the marketplace and plugin:

codex plugin marketplace add civictechdc/congressional-tech --ref congress-mcp-v0.4.0
codex plugin add congress-explorer@congressional-tech

Open the desktop app and start a local conversation. The button below prepares a question with Congress Explorer selected; review it and send when ready.

Explore Congress in Codex

Distributed through our repository marketplace. A public OpenAI directory listing requires separate approval. This server runs in local desktop conversations.

Add the research skill

Already using an agent with skills? Add our research and setup guidance with Node.js 22.20 or later:

npx skills add civictechdc/congressional-tech --skill explore-congress

Choose your agent, then ask it to use explore-congress. Add --global to make the skill available across projects.

This installs instructions only. Connect the MCP server using one of the setup paths on this page to query data. The full plugin already includes this skill. Skills CLI documentation ↗

01 / Install

Install Congress Explorer

Install uv, then paste this command into a terminal. It installs the released PyPI package and manages Python 3.12 for you.

uv tool install --python 3.12 congressional-explorer-mcp==0.4.0

If your shell cannot find the command afterward, run uv tool update-shell and open a new terminal. Windows is not yet supported.

02 / Prepare

Download the data and check it

congress-explorer-mcp prepare
congress-explorer-mcp check

The check should report "status": "ok" and six tools. It starts the server and reads a data row over MCP.

The October 30, 2026 snapshot is about 65 MiB. Data goes in ~/.cache/congress-explorer-mcp; Python and dependencies need additional space. Document files are fetched separately when your assistant needs them.

03 / Connect

Choose your assistant

Claude Desktop

congress-explorer-mcp config --client claude-desktop

Settings → Developer → Edit Config

Merge the printed congress entry into mcpServers in your configuration. Keep your existing servers. Quit and reopen Claude Desktop.

The command prints settings with the correct paths. It does not edit your client. Client documentation ↗

Codex

congress-explorer-mcp config --client codex

~/.codex/config.toml

Add the printed [mcp_servers.congress] section to your config file. Replace an existing congress section rather than adding a duplicate. Restart the client connection.

The command prints settings with the correct paths. It does not edit your client. Client documentation ↗

Cursor

congress-explorer-mcp config --client cursor

~/.cursor/mcp.json

Merge the printed congress entry into mcpServers in your configuration. Keep your existing servers. Enable Congress in Cursor’s MCP settings.

The command prints settings with the correct paths. It does not edit your client. Client documentation ↗

Claude Code

congress-explorer-mcp config --client claude-code

A new terminal session

Save the printed JSON to a file, then start Claude Code with --mcp-config /absolute/path/to/that-file.json. Review the requested tool permissions.

The command prints settings with the correct paths. It does not edit your client. Client documentation ↗

Start a conversation

Follow a question
through Congress.

Ask your assistant to use Congress Explorer. It can discover tables, follow documented relationships, and run read-only SQL.

“Find documents related to H.R. 1 in the 119th Congress. Group them by hearing.”

“Find hearings about artificial intelligence and list their witnesses and organizations.”

“Find recordings and saved transcripts for a committee meeting I choose.”

Know your data

Published records.
Clear boundaries.

What can I explore?

Meetings, witness appearances, bills, nominations, members, committees, rosters, recordings, and document records. All published tables and columns are queryable. Availability varies by source, Congress, and record; this is a bounded collection, not a complete record of every congressional activity.

Can it read documents?

The server searches structured data and document metadata, then returns links to saved files. Your assistant needs its own file-fetching and reading tools to read those documents. The server does not provide full-text search or OCR.

How do I get newer data?

With the plugin, ask your assistant to follow the setup skill's data refresh instructions. With a manual install, run congress-explorer-mcp prepare again, then restart the client connection. The extension uses the same default data directory, so a manual install can also refresh its data. Only changed files download. Software updates are separate: install the version linked on this page and restart your client.

What leaves my computer?

The server downloads public data during preparation and queries it locally afterward. Your AI client may send your questions and tool results to its model provider. Fetching linked documents requires an internet connection.

Need help connecting?

Run congress-explorer-mcp check in your terminal. If data is missing, run prepare first. If you chose a custom --data-dir, use that same directory for preparation, checks, and configuration. Report an issue with your operating system, client, and error message.

Technical documentation ↗Browse the data on the web →