Flowsearch MCP server
A Model Context Protocol server for AI assistants. It lets an assistant search the web with Flowsearch, search forums, and read Flowsearch's statistics, screenshots and API facts. It only reads; it changes nothing.
Connect
Use this as the server URL. The transport is streamable HTTP, and no API key is needed.
https://mcp.flowsearch.org/mcp
For clients that take a JSON configuration (Cursor, VS Code and others), this is usually enough:
{
"mcpServers": {
"flowsearch": { "type": "http", "url": "https://mcp.flowsearch.org/mcp" }
}
}
In Claude and ChatGPT, add it as a custom connector (MCP server) and paste the URL.
Tools (36)
search | Search Flowsearch, a from-scratch web search engine with its own crawler and index. |
llms_txt | Go to a site and read only its llms.txt, the short guide a site publishes for language models. Nothing else of the site is fetched. Set full to true for llms-full.txt, the long version. Find sites with llms_search first. |
llms_search | Search only the llms.txt files that Flowsearch has crawled. Give a site or topic, for example cloudflare. Nothing but llms.txt files is returned. |
shopping | Search the Amazon products Flowsearch holds (amazon.com and amazon.de) and get price, rating, brand and, for laptops, the screen size. Prices are as last read, not live. |
videos | Search the YouTube videos Flowsearch holds and get their title, address and description. |
stats | Live Flowsearch stats: pages indexed, sites, terms, crawl rate, the crawler's latest batches (new pages, storing speed), the GitHub crawl run, and the forums (posts per platform: Lemmy, Discourse, Flowforum; servers, communities, … |
amazon_overview | How many Amazon products Flowsearch holds (amazon.com and amazon.de), how many have a price, and the newest ones read. |
amazon_stats | The Amazon part of Flowsearch in numbers: products from amazon.com and amazon.de, how many have a price, and the newest products. |
api_endpoints | The endpoints of the Flowsearch API (version 1) with what each does, and the request limits (60 a minute for an address, 10,000 a minute with an API key). |
crawl_jobs | The forum and Amazon crawl on GitHub Actions, job by job (lemmy, discourse, languages, amazon, again): status, when each started and ended, and the step it is on now. Shows the latest run, or the last few runs. |
crawl_run | The GitHub Actions crawl: the latest runs of the Rust crawler with their state, the next seed batch, discovered sites and failures. |
crawl_runs | The last crawl runs of Flowsearch as the /live page lists them: when each started and ended, how many pages it added, and whether one is running now. |
crawler_batches | The crawler's most recent batches, newest first: when each was stored, how many pages it carried, how many were new, how long storing took, and the index total afterwards. |
data_dumps | The daily data dumps of Flowsearch (the pages added in the last 24 hours: address, site, title and description, as gzip JSON lines): file, number of pages, size, checksum and download address. |
forums_communities | The biggest forum communities in Flowsearch (Lemmy communities, Discourse categories) by number of posts. |
forums_languages | How many forum posts Flowsearch holds for each programming language (Python, Rust, Go, Julia, C / C++, Kotlin, Swift, Scala, Ruby, R, ...), from the Python and programming language forums. |
forums_newest | The newest forum posts the Flowsearch forum crawler has saved (Lemmy, Discourse, Flowforum): date, platform, instance, community, title and address. Can be limited to one kind of forum, including the Languages tab (Python and othe… |
forums_overview | How big the Flowsearch forums are: posts per platform (Lemmy, Discourse, Flowforum), servers, communities, the biggest forum sites, and what the forum crawl on GitHub Actions is doing now. |
forums_search | Search the Forums tab of Flowsearch: Lemmy, Discourse and Flowforum posts, searched by title, text, community and top comments, ranked like the Forums tab. Returns community, instance, score, comment count, date and the best comme… |
geocode | Find the latitude and longitude of a place with Flowsearch's own maps service. |
get_flowsearch_url | The public addresses of Flowsearch: the website, the API, its documentation, the MCP server, the data dumps and the account page. Give a query to also get the address of that search, or a site to get the address of its page. |
growth | How fast Flowsearch is growing: pages added in each of the last 10 minutes, 6 hours and 8 days (the charts of the /live page), plus the totals. |
recent_pages | The pages Flowsearch indexed in the last moments (what the /live page shows as it happens): title, site and address. |
roblox_overview | Flowsearch's Roblox data: how many games it holds, total players, and the most recently read games with creator, players now and visits. |
safety_check | Whether Flowsearch flags a website as unsafe (malware, phishing). |
search_by_looks | Find websites by how they look: describe a page's appearance in words, for example dark blue online shop. It searches the screenshot descriptions Flowsearch keeps. |
search_page | Search Flowsearch and choose the page and the page size (the plain search tool gives only the first 10 results). Returns up to 50 results a page, pages 1 to 10. |
search_site | Search inside one website with Flowsearch: the best pages of that site for a query. |
site_facts | What Flowsearch knows about one website: how many of its pages are indexed, how many addresses it has, and the addresses of its Flowsearch page, screenshot and search. |
site_pages | Pages Flowsearch holds from one website, optionally about a topic. Also tells roughly how many pages of that website are in the index. |
spell_check | Check the spelling of search words with Flowsearch's dictionary, and get the corrected search (for example pythn tutorail becomes python tutorial). |
suggest | Search-box suggestions: what Flowsearch completes a few typed letters to. |
top_sites | The websites with the most pages in the Flowsearch index. |
whats_new | Flowsearch's own news page: the newest changes and milestones, newest first. |
wiki_card | The short Wikipedia-style card Flowsearch shows for a name or topic: title, a few sentences and the source. |
youtube_count | How many YouTube videos Flowsearch holds for a topic, with a sample of titles (the videos tool gives full results). |
More
- flowsearch.org, the search engine itself
- API documentation (the JSON API behind several of these tools)
- Terms of use · Privacy · Impressum
- Source code: github.com/Luishae07/flowsearch-source (AGPL-3.0)
The content of the pages that the tools return belongs to its owners and authors.