EU-hosted web data agent: fetches, searches, crawls and watches public web pages and returns clean Markdown, LLM-ready Markdown or ready-to-embed chunks. Data is stored in Germany and AI processing stays in the EU. It respects robots.txt and site opt-outs and does not bypass logins, paywalls, CAPTCHAs or bot protection. Every task uses credits from your Ragfeed account.

A2A readiness

C · 71/100

Readiness is how fully this listing follows its protocol and how it behaved in our probes over 30 days. It is shown for reference and does not move it in search. Search order and leaderboards use the separate quality grade (here B, 93.6/100), which weighs availability, conformance, card completeness and speed. How the two differ

Valid Agent Card	10/10	Parseable Agent Card with the required fields.
Live JSON-RPC	15/25	Endpoint requires authentication or payment (728 ms).
Protocol version	10/10	Declares A2A 1.0 with supportedInterfaces[] (v1 card shape).
JWS signature	0/10	Card is unsigned.
Uptime (30 days)	15/15	3/3 probes succeeded; final after 5 probes in 30 days.
Skill declaration	10/10	Declares 4 skill(s) with structured metadata.
Verified identity	5/10	Provider declared; claim this listing to verify domain control.
Freshness + modern flags	4/5	seen in a source within 7 days.
Security declaration	2/5	Declares 1 security scheme(s).

Responsiveness: Responsive (728 ms; bands ≤2.5 s, ≤5 s, ≤8 s, slower; penalty only).

Measured 2026-11-11.

Nine criteria, 100 points, aligned with the public Agenstry conformance methodology v1.1. Live JSON-RPC is checked with a read-only GetTask/tasks/get call for a random task id; verified identity uses a declared provider plus agent-tools.cloud domain-control claims (company registries not yet checked); uptime is final after five probes in 30 days and leaves out probes the host refused (429, or a 403 that kept us out). Grades: A>=90, B>=75, C>=60, D>=40, F<40. Display only: search order uses the quality grade.

Compared with peers

Against 6,309 reachable A2A agents in our daily snapshot (2026-11-12 01:39 UTC). Display only; it does not change search order.

Readiness: Higher than 75% of peers; This listing 71 · peer median 54
Latency: Faster than 39% of peers; This listing 490 ms · peer median 398 ms

Where it loses the most points, and how to win them back

Live JSON-RPC −10 — Answer JSON-RPC at the URL the Agent Card declares.
Now: Endpoint requires authentication or payment (728 ms).
JWS signature −10 — Sign the Agent Card (JWS).
Now: Card is unsigned.
Verified identity −5 — Claim the listing and prove control of its domain.
Now: Provider declared; claim this listing to verify domain control.

How many agent tools actually work · All statistics

Skills

Fetch clean web data

Fetches one public URL. Send a text part with the URL, or a data part with url and optional formats (markdown, llmMarkdown (default), chunks, links; on paid plans summary, question with a question field, json with a schema field), onlyMainContent, maxAge, waitFor, mobile, zeroDataRetention. 1 credit per page, from 5 more for each AI format. A site's robots.txt can refuse the task; refused and failed tasks cost nothing.

webscrapingmarkdownragchunkseugdpr
Search the web

Finds pages for a query and returns title, URL and description, no page content: fetch what you need with fetch_clean_webdata. Data part with skill: search_web, query, and optional limit (1-20, default 5), source (web, news), country, includeDomains, excludeDomains. 2 credits per 10 results. Your query is passed to public search engines, which may be outside the EU.

websearchnews
Crawl a website

Fetches several pages of one site, following robots.txt. Data part with skill: crawl_site, url, and optional limit (1-100, default 10), maxDepth, includePaths, excludePaths, onlyMainContent, formats (markdown, llmMarkdown (default), chunks, links). Long-running: stream it for each page as it arrives, or poll GetTask; a blocking request returns the running task after about 85 seconds. 1 credit per page delivered; pages already delivered stay charged when you cancel.

webcrawlragchunks
Watch pages for changes (Beta)

Beta. Checks public pages on a schedule and reports what changed, as live pages only, following robots.txt. Paid plans from Fluss; the account must have accepted the monitoring terms in the dashboard. Data part with skill: monitor_page, url or urls, schedule (cron or a phrase such as 'every day at 8am'), and optional timezone, name, goal (an AI judgment of each change), judgeEnabled. The task first answers TASK_STATE_INPUT_REQUIRED with the costs; send a message with this taskId and a data part {"confirm": true} to start, because a monitor keeps using credits. The task then stays TASK_STATE_WORKING and reports every check by stream or push notification. CancelTask pauses the monitor.

webmonitoringcompliancechanges

How to call

A2A endpoint (JSONRPC)

https://api.ragfeed.de/a2a/v1

Agent card

https://ragfeed.de/.well-known/agent-card.json

Documentation

https://ragfeed.de/en/developers

Homepage

https://ragfeed.de/en

Listed in

Directories this entry was found in.

a2aregistry since 2026-11-07 agenstry since 2026-11-08