Find 1920s writing that mentions the Tulsa race massacre and show me the page.
search_full_text, search_inside, get_hocr, get_page_imageget started
It runs on your own computer, as a program your AI client starts, and only reads unless you start it in write mode.
coming soon The first public release isn't out yet: the package isn't on PyPI and the repository is private, so the commands on this page won't work until then. They are here so you can see what setup will involve.
The server is a Python program run by uv, which fetches it and its dependencies. You need uv and Python 3.10 or newer.
claude mcp add internet-archive -- uvx internet-archive-mcp
{
"mcpServers": {
"internet-archive": {
"command": "uvx",
"args": [
"internet-archive-mcp"
]
}
}
}Any MCP client that can start a local (stdio) server takes the same command. The analysis tools need DuckDB: use uvx --from 'internet-archive-mcp[analysis]' internet-archive-mcp to include it.
claude plugin install internet-archive --marketplace eldridgeh/internet-archive-mcp
It adds the server and a short skill on using the Archive well: which tool to use, rate limits, untrusted content and the write rules. It starts in read mode.
No account or keys are needed to search and read. Ask in plain words and the assistant picks the tools. Some requests that were run against archive.org on 31 October 2026 for the project's examples:
Find 1920s writing that mentions the Tulsa race massacre and show me the page.
search_full_text, search_inside, get_hocr, get_page_imageWhat did example.com look like in 2005, and what has changed since?
wayback_history_summary, wayback_fetch, wayback_compareCan I borrow The Hunger Games? The ISBN is 9780439023528.
book_availability, openlibrary_editionWhat does the narrator say a minute into Duck and Cover?
search_items, get_item, get_subtitlesHow did TV news cover the April 2024 solar eclipse? Quote a clip.
tvnews_program_counts, tvnews_search, tvnews_captionsGive me a picture of the Prelinger Archives.
search_summary, search_items, item_viewsWatch whole runs on Ask the Archive and in the stories. account_info says which mode the server is in and whether keys are set up.
Uploads, edits and a few reads need your archive.org keys, an S3 access key and secret key. Copy them from archive.org/account/s3.php (signed in to archive.org in your browser), then enter them yourself:
uvx internet-archive-mcp configure
It asks for each key without showing it, checks it with archive.org and saves it to the ia.ini file the server reads. The plugin also takes keys in its settings. Never paste keys into a chat: no tool takes them, and the server never returns or logs them. Restart the server after setting them.
Start the server with --mode write to let the assistant upload and edit with your account (11 more tools), and add --upload-root to name the one folder it may upload from. Every write is previewed first and runs only when the assistant calls again with the preview's id; what it does is public and done as you, so start small, with test_collection. More on safety.