Georgia Commons

For anyone who uses an AI assistant

Bring Georgia's laws, bills, and court opinions into Claude and ChatGPT

A connector lets your assistant read the documents on this site directly, instead of relying on whatever it happens to remember. Add one address once, then ask your questions in the app you already use.

Machine-readable https://georgiacommons.org/mcp.md · https://georgiacommons.org/index.md · MCP https://mcp.georgiacommons.org/mcp

What it is

A connector, usually called an MCP server (Model Context Protocol), is a small bridge between an AI assistant and a real source of information. Georgia Commons runs one. Once it is connected, an assistant that is asked about a Georgia statute, a bill, or a Supreme Court decision looks the document up and answers from its actual text, with a link to the page it read.

One address covers the Code and the two constitutions, bills, and opinions. It is read-only, it needs no account, and there is nothing to pay for on this end.

How to connect it

In every app the step is the same: paste this address.

Connector URL

https://mcp.georgiacommons.org/mcp

There is no sign-in, no password, and no key. Menu names in these apps change fairly often, so treat the steps below as the shape of the thing rather than the exact words on your screen.

Claude, the easiest

Add it once and it follows you to every Claude app

  1. Open Settings, then Connectors.
  2. Choose Add custom connector.
  3. Give it a Name. "Georgia Commons" works.
  4. Paste the address above into Remote MCP server URL. Leave the advanced and OAuth fields alone.
  5. Click Add. In a chat, open the tools menu and switch it on.

This works on the Claude web app, the desktop app, and the phone app, and a connector added in one shows up in the others. Every plan including Free can add at least one custom connector. On a Team or Enterprise plan an owner adds it once under organization settings and everyone else turns it on.

ChatGPT

Add it from the website, not the phone app

  1. Open ChatGPT on the web. The phone and desktop apps do not offer this yet.
  2. Go to Settings, then Apps, and choose Create app.
  3. Enter a Name. "Georgia Commons" works.
  4. Leave Connection on Server URL and paste the address above.
  5. Set Authentication to No Auth, confirm, then click Connect.

Deep research and the Responses API reach the same server through its search and fetch tools, so anything connected this way can cite the page it read.

Claude Code and other terminal agents

One command

claude mcp add --transport http georgiacommons https://mcp.georgiacommons.org/mcp

Any client that speaks the streamable HTTP transport can point at the same address. There is nothing else to configure.

OpenAI Responses API

From your own code

from openai import OpenAI

client = OpenAI()
response = client.responses.create(
    model="gpt-5",
    tools=[{"type": "mcp", "server_label": "georgiacommons",
            "server_url": "https://mcp.georgiacommons.org/mcp", "require_approval": "never"}],
    input="What does Georgia law require a landlord to do with a security deposit? Quote it and cite it.",
)
print(response.output_text)

What to ask once it is connected

  • "What does Georgia law require a landlord to do with a security deposit? Quote the statute."
  • "What does House Bill 136 of the 2025-2026 session actually do, and where does it stand?"
  • "Summarize the Supreme Court of Georgia's opinion in docket S26A0060 and quote the holding."
  • "Which chapter of the Georgia Code covers landlord and tenant law? List its sections."
  • "Is this sentence really in O.C.G.A. 44-7-34?" and then paste the sentence.

One tip that changes the results more than anything else: Search matches the words of the document. Ask in the words the document would use, not the words of the question. A statute about a child says "minor"; an opinion about a car crash says "motor vehicle". If the first search comes back thin, ask again in the document's words.

Why this beats asking an assistant on its own

  • It reads the document. Answers come from the stored source text, not from what the model absorbed in training or found in a search result.
  • Quotes can be checked. The connector has a tool that reports whether a quotation appears word for word in the stored text. Ask it to check its own quotes and it will.
  • The Code carries no interpretation. The statutes and both constitutions are source text only. Georgia Commons does not summarize, explain, or interpret them anywhere, including here.
  • It links to the original. Every answer can name the page it read, and every page links to the official document the state, the General Assembly, or the court published.
  • It says when it does not know. Ask it what it covers and it reports what is actually loaded at that moment rather than guessing.

Good to know

  • It only reads. Every tool is read-only. The connector changes nothing, and Georgia Commons does not receive or store your conversations.
  • No account, no key. The documents are public, so the connector is too. Your assistant's own plan is the only thing that might limit you.
  • Not a government service. Georgia Commons is an independent project of Georgia Civic Data. It is not affiliated with, operated by, or endorsed by the State of Georgia, its courts, or any agency.
  • Not legal advice. It hands your assistant the text of the law. Reading a statute is not the same as knowing how it applies to you.
  • Assistants still make mistakes. A connector removes the guessing about what a document says, not the risk that an answer overstates it. Follow the link for anything that matters.

What the connector can do

The server offers 23 tools. You never call them yourself; your assistant picks the one it needs. They are listed here so you can see exactly what the connector is allowed to do before you add it.

Across the whole site

Five tools that work over every collection at once. The first two are the cheap orienting calls an assistant makes before it goes looking for anything.

  • `describe_site` What Georgia Commons holds, which collections this server has loaded, how identifiers are written, and which tool to call next.
  • `get_coverage` Exactly what is loaded right now, read from the stores at the moment of the call: coverage sentences, corpus versions, counts, sessions, years, and licenses.
  • `search` Search every collection, or one, and get back a short list of matches with a title and a link. This is the tool ChatGPT's connectors and deep research use.
  • `fetch` Read one document by its identifier and get the same text the site serves, with the page address to cite.
  • `verify_quote` Check whether a quotation appears word for word in the stored text. Deterministic, with no model involved.

The Code and the two constitutions

The Official Code of Georgia Annotated and the Georgia and United States constitutions, as source text. No summary, no interpretation, no model anywhere in this path.

  • `code_describe_structure` How the Code is organized and how citations are written, plus the list of titles. The cheap first call for statute questions.
  • `code_lookup_section` One section by its citation, written any of the ways people write it.
  • `code_lookup_sections` Several sections at once, as a list or a range such as 44-7-30 through 44-7-35.
  • `code_browse` Walk the hierarchy: the titles, a title's chapters, or a chapter's sections.
  • `code_get_chapter` Every section of one chapter with its text, a window at a time.
  • `code_search` Full-text search of the Code. A citation among the words returns that section first, and an act's short title returns the section that states it.
  • `code_find_chapter` For when the topic is known but the citation is not: the titles and chapters whose headings carry the words.
  • `code_find_act` Laws by the short title the statute itself states, such as Joshua's Law.
  • `constitution_lookup` One provision of the Georgia or United States constitution as printed in the Code.
  • `constitution_browse` The articles and sections of either constitution, one line each.

General Assembly bills

Bills of the Georgia General Assembly: the full text as filed, and the plain-language summaries Georgia Commons wrote, always labeled and always after the text.

  • `bills_lookup` One bill by its number, written any of the ways people write it.
  • `bills_get_text` The stored bill text alone, a window at a time.
  • `bills_list` The bills of a session in number order, filtered by chamber, kind, status, or subject.
  • `bills_search` Full-text search over bill text, titles, and labeled summaries.

Supreme Court of Georgia opinions

Opinions of the Supreme Court of Georgia, with the same rule: the court's text first, anything Georgia Commons wrote after it and labeled.

  • `opinions_lookup` One opinion by its docket number, reporter citation, CourtListener id, or case name.
  • `opinions_get_text` The stored opinion text alone, a window at a time.
  • `opinions_list` The opinions filed in a year, in filing order, filtered by case type.
  • `opinions_search` Full-text search over opinion text, case names, and labeled summaries.

If your assistant cannot add a connector

Every document page here has a small row of buttons under its title that does much the same job one page at a time.

  • Copy for LLM. Puts the page's full source text on your clipboard, ready to paste into any assistant.
  • Send to AI. Opens Claude or ChatGPT with the page's address already in the prompt, so it can fetch the document on its first turn.
  • View as markdown. Every page has a plain Markdown version at its address plus .md, and the source text alone at its address plus /text.

Building something rather than chatting? The JSON API needs no key either, and https://georgiacommons.org/for-assistants is the one-screen reference for an agent's operator: the URL grammar, one worked example per collection, and every address written out in full.