RagfeedLLM-ready web data. Firecrawl-compatible, hosted in Germany.
Ragfeed turns web pages and documents into clean Markdown, structured JSON or ready-to-embed chunks. The API speaks Firecrawl v1 and v2, so your existing code and the official SDKs keep working – you swap the base URL and the key. Data is stored in Germany, AI processing stays in the EU.
Median, our measurement on 20 German-language pages, 1 Oct 2026. What the numbers mean · Method and per-page results
Where do you want to start?
Developers First request in three steps curl, firecrawl-py and Node with one changed line, the formats llmMarkdown and chunks, limits. Developer quickstart → Claude and MCP clients Use it from Claude Remote MCP server with scrape, search, map, crawl and extract. Sign-in by e-mail code, no API key needed. MCP connector →Three reasons to try it
Compatible with the code you have, run in Europe, and built for what a language model actually needs from a page.
Compatible
Firecrawl API v1 and v2. The official firecrawl-py and firecrawl-js SDKs run unchanged against Ragfeed in our
test suite; the official Firecrawl n8n node works with the base URL https://api.ragfeed.de/v2. Not supported today: the formats
branding, audio and video, and persistent browser sessions.
European
Servers in Germany. AI formats are processed by an EU provider that does not store requests or train on them; results are deleted after a short period. Data processing agreement (DPA) on request in every plan, including the free one. Prices in euros. Details (German): data locations.
Made for LLMs
Two formats only Ragfeed has: llmMarkdown (main content with a short source header) and chunks (the page split
along its headings into records for your vector database – id, heading path, token count, hash). No extra credits for either.
What the token numbers mean
Both numbers are medians over 20 pages (6 start pages, 14 articles),
tokenizer o200k_base, measured on 1 Oct 2026. Against the full page as text the saving ranges from
18 % to 90 %; pages that are already lean save little.
Most of the 63 % comes from main-content extraction – dropping menus, footers and cookie notices – which other
extractors do as well; the extra cleanup in llmMarkdown adds a median ~9 %. So read the
63 % as “main content vs. whole page”, not as a comparison with another product.
The pages are German-language sites (Wikipedia, MDN, news, government, documentation, recipes). Full write-up with every page: raw HTML vs. full-page text vs. main-content Markdown.
Use it from Claude (MCP)
Ragfeed is listed in Claude's connector directory. The remote MCP server at https://api.ragfeed.de/mcp offers scrape,
search, map, crawl and extract tools. Sign in via OAuth 2.1 with an e-mail code – no API key needed – or send your API key as a Bearer token.
It is also in the official MCP Registry (io.github.LeifAlbrand/ragfeed), on Smithery and on Glama.
Claude Code:
claude mcp add --transport http ragfeed https://api.ragfeed.de/mcp
Usage is billed in your Ragfeed credits, same as the API.
Pricing
1 credit per page (Markdown, llmMarkdown, chunks, HTML, links, screenshot); AI formats +5 credits.
Failed pages cost nothing.
| Plan | Price / month | Credits / month | Concurrent requests | Requests / min |
|---|---|---|---|---|
| Quelle (free) | €0 | 1,000 | 2 | 10 |
| Bach | €9 | 5,000 | 5 | 60 |
| Fluss | €49 | 100,000 | 15 | 300 |
| Strom | €199 | 500,000 | 30 | 600 |
The free plan needs no credit card. Paid plans are monthly and can be cancelled at any time; extra credit packs never expire. Prices are final prices – no VAT is charged (German small-business rule, § 19 UStG). Checkout and invoices are in German.
Honest limits
- No SLA. Ragfeed is run by its founder in Germany; you talk to the person who built it.
- Capacity is limited by plan (2 / 5 / 15 / 30 concurrent requests). If you need more, write before you hit the limit.
- Large platforms and marketplaces often block data-centre traffic. We do not work around that, and we do not bypass logins, paywalls or CAPTCHAs. On documentation, articles, company and government sites the success rate is above 90 %.
robots.txtis respected for crawl and map.
In German only
Sign-up, dashboard, checkout, e-mails and legal texts are in German. Your browser can translate them; the API responses, error codes and SDKs are language-neutral. Questions in English are welcome: contact or kontakt@ragfeed.de.