One secure address gives ChatGPT or Claude read-only access to Living Inflation Data's consumer-price series, observations, coverage and revision history. No invitation is needed: self-service access accepts any verified email address. Read limits are generous for research questions and charts, while preventing bulk extraction of the database.
https://mcp.iemlab.org/mcp
Set this up in ChatGPT on the web. Open chatgpt.com. Custom MCP access depends on your account and workspace policies; ask your workspace admin if the option is missing.
For Team and Enterprise workspaces, an authorized owner or admin must first add the connector under Organization settings → Connectors. Members then select Connect individually. Claude setup guide
Instructions checked October 14, 2026. Menus can vary by account and rollout.
Use the full https://mcp.iemlab.org/mcp address, including /mcp.
Hosted MCP requires email verification; no invitation or API key is needed.
If a saved connection stops working, reconnect and complete sign-in again.
Codex and Claude Code can connect directly to the hosted server without creating a ChatGPT plugin or a Claude web connector. This works when your organization permits custom MCP in that client. A restriction in the chat product alone does not establish whether the coding client allows it; managed MCP restrictions still apply.
You need the relevant command-line client installed and signed in, plus permission to use it with external data services. These commands connect to Living Inflation Data's hosted database; you do not need to clone this project, run a local server, or supply an API key for Living Inflation Data.
In the desktop app, open Settings → MCP servers → Add server, choose Streamable HTTP, and enter the address above. Save, restart when prompted, then select Authenticate.
Alternatively, run these commands in a terminal on the machine where you use Codex:
codex mcp add living-cpi-data --url https://mcp.iemlab.org/mcp
codex mcp login living-cpi-data
codex mcp list
Complete the browser email sign-in opened by the login command. Start a new
Codex session and use /mcp in the CLI to check the connection,
then ask a question below. Codex shares MCP configuration across its local
clients. Codex MCP guide.
From the project folder where you want to use the data, run:
claude mcp add --transport http living-cpi-data https://mcp.iemlab.org/mcp
claude mcp list
Start Claude Code in that folder, enter /mcp, select
living-cpi-data, and complete the browser authentication.
This default registration is private to you in the current project.
Claude Code MCP guide.
Ask your administrator to approve https://mcp.iemlab.org/mcp
as an OAuth-authenticated, read-only external data service, or to provide
a managed connection. Local configuration cannot override a managed block.
You can also browse the country pages, select the series and dates you need, and download a CSV. Where your workspace permits file uploads, attach that CSV to a chat or open it in your coding project and ask the assistant to analyze it. This uses a downloaded snapshot; the assistant cannot query new data through the file.
Compare food and housing inflation in Pakistan's urban and rural CPI baskets, including their current basket weights.
Compare Japan's official Statistics Bureau CPI with the IMF series over the last five years. Flag disagreements and keep each published base separate.
Find an important-country CPI gap that remains unexplained, and contrast it with an explained gap such as the U.S. government shutdown.
Living Inflation Data's tools only read the published replica. They cannot modify the database, its sources, or your account. Responses are bounded to twenty series and 10,000 observations; rolling allowances are shared with authenticated website downloads. Ordinary plotting uses Living Inflation Data's fast server-rendered image; ask explicitly for an interactive chart only when you need hover or toggles.
Living Inflation Data applies usage limits to keep public access reliable. Anonymous website browsing requires no account; hosted MCP and larger website downloads require a verified email. Operational records are limited to what is needed to enforce those limits, diagnose failures, and measure aggregate demand. The usage record does not store your question text, returned data, raw IP address, or email address, and activity is not used to profile research topics.