apify-mcp-server review
TestedExtract data from any website with thousands of scrapers, crawlers, and automations on Apify Store ⚡ · directory profile
Broad pass rate through a two-step actor flow — at double-digit latencies
Running the RAG Web Browser actor through Apify's MCP server passed 6 of 7 applicable tasks — every search and every server-rendered scrape. The cost of that coverage is latency: each task is an actor run followed by a dataset fetch, and completed tasks ranged from 9.5 to 25 seconds with a 15.7-second median, the slowest in the run. The JavaScript-rendered fixture was its one failure.
Measured strengths
- Passed all three search tasks — one of only three services to do so in this run
- All server-rendered scrapes passed
- Usage-based platform billing rather than per-request pricing can suit bursty workloads
Measured caveats
- Slowest service in the run: 9.5–25 s per completed task (p50 15.7 s, p95 22.8 s)
- Each benchmark task required two tool calls (actor run + dataset fetch) — reflected in our cost accounting
- Failed the JavaScript-rendered scrape in this run — the fetched content lacked the client-side-rendered portion
- No structured-extraction task type
- No fixed per-call price exists, so our cost figure is a documented estimate, not metered usage
Pricing
The RAG Web Browser actor is free to install; usage bills Apify platform compute units (1 CU = 1 GB-hour of actor runtime) per the actor's listing (apify.com/apify/rag-web-browser, checked 2026-07-17). There is no fixed per-call rate — our per-task cost is a labeled estimate.
Benchmark results
Run 20260717T113200Z-7113 · promoted 2026-07-17 · harness 88581d0 · how these numbers are produced
- Success rate
- 86% (7 applicable tasks)
- Latency p50 / p95
- 15737 ms / 22809 ms
- Cost per task (estimated)
- ~$0.004
Capability coverage
-
scrape_markdown: supported -
web_search: supported -
structured_extract: not offered (excluded from scoring)
Task-by-task outcomes
| Task | Outcome | Latency |
|---|---|---|
| Scrape a minimal static page | passed | 17046 ms |
| Scrape a long-form encyclopedia article | passed | 14866 ms |
| Scrape a technical documentation page | passed | 14371 ms |
| Scrape a JavaScript-rendered page | failed Returned content without the client-side-rendered portion | 9453 ms |
| Known-answer search (historically stable fact) | passed | 17846 ms |
| Known-answer search (protocol constant) | passed | 15737 ms |
| Navigational search for a canonical domain | passed | 24936 ms |
| Extract a titled summary as structured JSON | not supported | — |