String for AI agents: benchmark results
String (scraping infrastructure) was tested on scrape, search and map, not on crawl. It got the same tasks, the same model (Claude Haiku 5.5), the same prompt and the same limits as every other provider, and a script checked every answer.
Scrape 40 tasks · Test heading and structure preservation on documentation pages, plus package versions and JS fields.
- Tasks done
- 88%
- 35 of 40
- 95% interval
- 76%–97%
- Cost per finished task
- 1.33¢
- Median time
- 3.9 s
4th of 10 on tasks done. String is behind Context.dev (95%), but the gap of 7 points is too small to rule out chance.
Answer in the provider's response, no model: what the API returned held every answer on 38 of 40 applicable tasks (95%).
| Provider | Tasks done | 95% interval | Cost per finished task | Median time |
|---|---|---|---|---|
| Context.dev | 38 of 40 (95%) | 85%–100% | 1.36¢ | 4.6 s |
| Spider | 37 of 40 (93%) | 81%–100% | 0.88¢ | 5.7 s |
| Scrapfly | 36 of 40 (90%) | 79%–100% | 1.17¢ | 6.0 s |
| String | 35 of 40 (88%) | 76%–97% | 1.33¢ | 3.9 s |
| Linkup | 34 of 40 (85%) | 71%–96% | 1.23¢ | 4.0 s |
| Firecrawl | 32 of 40 (80%) | 66%–92% | 1.67¢ | 4.8 s |
| TinyFish | 31 of 40 (78%) | 61%–92% | 0.71¢ | 4.6 s |
| Browserbase | 30 of 40 (75%) | 60%–89% | 1.93¢ | 3.6 s |
| You.com | 29 of 40 (73%) | 56%–89% | 1.00¢ | 3.6 s |
| Jina Reader | 6 of 40 (15%) | 5%–26% | 4.26¢ | 5.4 s |
Search 30 tasks · Find facts, registry versions and documentation headings on the live web.
- Tasks done
- 77%
- 23 of 30
- 95% interval
- 56%–96%
- Cost per finished task
- 1.32¢
- Median time
- 6.8 s
5th of 7 on tasks done. String finished 17 points fewer tasks than Linkup, the leader. That gap is big enough to rule out chance.
Answer in the provider's response, no model: what the API returned held every answer on 19 of 27 applicable tasks (70%).
| Provider | Tasks done | 95% interval | Cost per finished task | Median time |
|---|---|---|---|---|
| Linkup | 28 of 30 (93%) | 85%–100% | 2.19¢ | 6.2 s |
| Firecrawl | 26 of 30 (87%) | 71%–97% | 2.91¢ | 6.9 s |
| TinyFish | 26 of 30 (87%) | 70%–100% | 0.97¢ | 8.2 s |
| Context.dev | 25 of 30 (83%) | 63%–100% | 1.39¢ | 8.4 s |
| String | 23 of 30 (77%) | 56%–96% | 1.32¢ | 6.8 s |
| Browserbase | 21 of 30 (70%) | 50%–87% | 2.79¢ | 15 s |
| You.com | 21 of 30 (70%) | 50%–87% | 2.64¢ | 4.8 s |
Map 15 tasks · Discover canonical URLs within a verified documentation scope.
- Tasks done
- 93%
- 14 of 15
- 95% interval
- 80%–100%
- Cost per finished task
- 0.98¢
- Median time
- 8.4 s
1st of 3 on tasks done. String finished the most map tasks, but its lead over Firecrawl (87%) is too small to rule out chance.
Answer in the provider's response, no model: what the API returned held every answer on 14 of 15 applicable tasks (93%).
| Provider | Tasks done | 95% interval | Cost per finished task | Median time |
|---|---|---|---|---|
| String | 14 of 15 (93%) | 80%–100% | 0.98¢ | 8.4 s |
| Firecrawl | 13 of 15 (87%) | 67%–100% | 1.32¢ | 11 s |
| Context.dev | 12 of 15 (80%) | 60%–100% | 0.98¢ | 6.4 s |
String compared
- String vs Context.dev scrape, search and map, compared with Context.dev
- String vs Firecrawl scrape, search and map, compared with Firecrawl
- String vs Spider scrape, compared with Spider
- String vs Scrapfly scrape, compared with Scrapfly
- String vs Linkup scrape and search, compared with Linkup