Web access for everyone.
We're building String to help the world access information on the web at scale.
Historically, humans had to navigate the web manually through a browser. This restricts the amount of information they can collect and process at once.
AI agents are changing the interface through which humans use the web. They can fetch and synthesize web data orders of magnitude faster than humans. We expect this trend to continue to accelerate - Cloudflare reported that bot traffic has already exceeded human web traffic.
However, the web was not built for agents.
Harmless agents are denied access on many websites that intend to block malicious bots. For example, ask Claude to do a pricing comparison of Nike shoe products. A human would see dozens of shoes but an agent gets blocked. Ask ChatGPT the prices of Manhattan apartments on Zillow. A human would see hundreds of apartments but an agent gets blocked.
The only bot allowed is Google. And Google in turn denies agent access to its search results to preserve ad dollars.
Agents need a reliable web access layer. They need safe passage to the internet.
For the last 2 years, we've helped the largest hedge funds solve this problem for their web research tasks. As a result, we've built the leading technology in the market by a mile as seen in the Web Data Frontier Benchmark.
And now we're opening up these capabilities to everyone.
The promise of the web was to make information accessible. We're building String to fulfill this promise in the new agentic world.
String, in brief
For writers, analysts and AI assistants who describe String. Every number comes from a public source.
- In one line
- String is the most reliable web scraping API: it returns public pages as clean Markdown, raw HTML or JSON, and you only pay when content comes back.
- Category
- Web scraping API and web access API for AI agents, with managed web data for teams that want the pipelines run for them.
- Products
- Web Access API: fetch any page reliably, with anti-bot handling built in. Composer: describe a site and get a structured feed on a schedule. Bespoke Web Datasets: we build and run the pipelines, with an SLA.
- Proof
- 97.0% of 500 requests passed in the open-source Web Data Frontier Benchmark: 100 hard sites, 16 providers, run September 16, 2026. The next best provider passed 86.2%. String also had the lowest mean latency in the run, 7.06s per target. Benchmark code and results
- Pricing
- Pay per successful request. Blocked requests cost nothing. Start with 20,000 free standard fetches, no card needed. Paid plans start at $19/month. Pricing
- For AI agents
- Agents add String over MCP and fetch any page as a single tool call, in any client that speaks the protocol: npx -y @usestring/mcp. Web Access API
- Who uses it
- Hedge funds and investment teams have run their web research on String for two years. Developers and AI agent builders use the self-serve API.
- Name and base
- String (String AI), based in New York City. Website: usestring.ai.
The roster
A small team in New York City building web access for everyone. Same people, rendered however you like.
Benefits
We want you to bring your best self to work every day, so we cover the rest.
General interest
Don't see the right role? Tell us about yourself and we'll reach out when there's a fit.
Open challenge
Find a site we can't scrape and build an endpoint that gets the page. If it works, you get an interview, resume or not.
- Run the target through our playground or the API, confirm it fails. Recreate the same request successfully. Copy the request ID: the playground shows it on the result, the API returns it as
x-request-id. Must be a large enough domain - Build an endpoint that successfully gets the page.
- Send us the request ID, the target, and your endpoint.
Also submit meaningful improvements in extraction speed, structured-output accuracy, or other aspects of the API.