Web access for everyone.

Never get blocked again - the best web data API ever made.

Fetch any URL

POST /v1/fetch

▍

Try:

Trusted by dozens of enterprises

6 of top 10 hedge fundsUnicorn startups2 of top 5 asset managers

“Our product depends on reliable web data. Before String, we were constantly getting blocked and spending countless engineering hours debugging. The API just works and they've become a core part of our infra stack.”

Danny KellerCTO at Revenue Vessel

Built for coders and vibe coders.

Four lines of code or one sentence to your agent.

Your scraper, before and after

scraper.py

import requests, time, random

from bs4 import BeautifulSoup

from itertools import cycle

PROXIES = cycle(open("proxies.txt").read().split())

UA_POOL = load_user_agents("agents.json")

def fetch(url, retries=5):

for i in range(retries):

try:

r = requests.get(url, proxies=next(PROXIES),

headers={"User-Agent": random.choice(UA_POOL)})

if r.status_code in (403, 429):

rotate_fingerprint(); continue

return BeautifulSoup(r.text, "lxml")

except Exception:

time.sleep(2 ** i + random.random())

raise RuntimeError("blocked. again.")

import requests

r = requests.post("https://request.usestring.ai/v1/fetch",

headers={"Authorization": "Bearer sk_live_…"},

json={"url": "lowes.com/c/drills", "format": "markdown"})

# that's it - never blocked

Your agent, with String

✳Claudenew chat

Scrape today's prices for every DEWALT drill on lowes.com

I'll grab today's DEWALT drill prices from lowes.com. Trying a direct request first, then a headless browser…

✗ GET lowes.com/c/drills → 403 Forbidden · 5/5 retries failed

I wasn't able to get through - lowes.com's bot protection is blocking every request, even from the headless browser. I don't have a way past it on my own.

Try again - use the String API

✓ POST request.usestring.ai/v1/fetch → 200 OK · 1.2s

That worked. I pulled all 42 pages and extracted 1,284 prices as clean markdown. Want me to save them to a CSV?

Best in class, by 15 points.

Aggregate success across 495 requests per providerRun the open-source Web Data Frontier Benchmark on GitHub →Why we built our own benchmark →

Q3 2026 · Web Data Frontier Benchmark

Updated monthly.

Faster.
More accurate.
More reliable.

99 hard targets
495 requests per provider
14 of 15 tested providers shown, head to head

See full methodology on GitHub↗

Success rate is the share of requests that returned the expected content.

Success rate

string

97.0%

context_dev

79.2%

firecrawl

78.6%

bright

78.0%

oxylabs

76.8%

zyte

72.7%

decodo

70.7%

nimble

66.3%

scraperapi

64.2%

scrapingdog

54.3%

browserbase

42.2%

zenrows

34.3%

scrapingant

30.7%

scrapingbee

29.7%

0%25%50%75%100%

Everything on the web, given to you

We meet you where you're at - whether you want a simple API, DIY web data pipelines, or datasets delivered to you.

Live · fetch

POST /v1/fetch

url "zillow.com/homedetails"

200· 0.41s

# 4-bed home · Austin, TX

$540,000 · 3,200 sqft · 4d on market

clean markdown

Web Access API

Access any web data

Search, scrape, and map any site with our state of the art API.

Use in Claude Use in ChatGPT

String Composer

Track all the running shoes on nike.com and refresh every day

Schema drafted: name · price · sizes

Source mapped: nike.com/w/running-shoes

Schedule set: daily · 6:00am

building feed · nike.com

String ComposerComing soon

End-to-end pipelines

Build production grade web data pipelines with data quality monitoring via just a prompt.

Forward deployed feed

DatasetCadence

Retail pricing & promosDaily

Store locationsWeekly

Telecom availabilityDaily

SEC & company filingsRealtime

SLA 99.9% · to your warehouse

Forward deployed

Bespoke Web Datasets

Our FDEs work with you to build and maintain scaled web data pipelines.

Book demo

We integrate with whatever you do.

Reach String programmatically, through MCP, or have data delivered straight into the tools your team already uses.

Use the Web Access MCP server to fetch any page on the web, live and structured. Paste this into Claude, Cursor, or any client with a config file:

claude_desktop_config.json

{
    "mcpServers": {
        "string-ai": {
            "command": "npx",
            "args": ["-y", "@usestring/mcp"],
            "env": { "STRING_AI_API_KEY": "sk_live_…" }
        }
    }
}

Claude.ai → Settings → Connectors → Add Custom Connector
Perplexity → Settings → Connectors → + Custom connector
Paste this URL with your API key in the path

https://mcp.usestring.ai/YOUR_API_KEY/v1/mcp

In Perplexity, pick Remote, set Transport to Streamable HTTP, and leave Authentication on None: the key in the URL is the credential. Custom connectors need a paid Perplexity plan.

Add to Claude ↗

Or delivered into your stack

Reach it the way your team already works, then land it wherever your data lives.

REST API

Your services, any language

MCP server

Claude, Cursor, ChatGPT, Perplexity, any agent

Automations

Airflow · Zapier · n8n · Make

Warehouse sync

BigQuery · Snowflake · Databricks · Postgres · Redshift · ClickHouse · S3