Integuru builds fast integrations with platforms that lack official APIs: its agent watches a user's browser session, reconstructs the internal request graph and emits code that talks to the platform's own endpoints; the company charges for custom builds, hosting and authentication management.

That LLMs can decipher platforms' internal API patterns from captured browser traffic and write runnable integration code, so no official API is needed.

Richard Zhang and Alan Lu, recent college graduates, wanted to make US income-tax data accessible; banks, brokerages and payroll companies ignored their requests for API access, so they reverse-engineered those systems to pull W-2s and 1099s. Browser automation proved unreliable and slow, so after manually reverse-engineering dozens of platforms and spotting patterns LLMs could decipher, they built an agent to automate the work and open-sourced it as Integuru (YC W24).

The v0 repo takes a HAR file of browser requests, cookies and a plain-language prompt — for example 'download utility bills' — and outputs runnable Python that hits the platform's internal endpoints. In the launch thread the team said Integuru began as an internal tool and was expanding beyond the tax industry, charging for custom integrations, hosting and cookie-based authentication; commenters pressed on what happens when unofficial APIs change and where the legal line sits, and the founders answered with auto-patching plans and a focus on use cases where users own their data.

Traction on named platforms: Launch HN on 2024-10-29 (item 41983409) drew 218 points and 83 comments, and the GitHub repo held 4.8k stars and 381 forks at the 2026-09-04 crawl.

A clear, falsifiable technical bet: that LLMs can read internal API patterns from captured traffic and emit working integration code — replacing the humans who reverse-engineer by hand.

A visible origin in the launch text: refused API access led the founders to build integrations anyway, and the pain of doing that manually is the wedge the company still sells against.

A live debate in the thread — breakage and CFAA or ToS risk — that defines the bet's ceiling, with the founders answering with auto-patching and user-owned data.

When vendors refuse you API access, the reverse-engineering you do anyway can become the product — if you can automate it and sell reliability, with breakage and legal risk as the ceiling.

As of 2026-09-05 Integuru is still building: the GitHub repo now labels v0 'the earliest version we released publicly' and points to a newer product at integuru.com, whose site claims 10M+ monthly API calls, YC backing and paid plans from free to $499+/month. No funding amount or exit appears in the record; direct HTTP integration instead of browser automation remains the product's stated core.

FOLLOW THE EVIDENCE

The sources

  1. Launch HN: Integuru (YC W24) – Reverse-engineer internal APIs using LLMs (218 points, 83 comments) Hacker News · news.ycombinator.com ·
  2. Integuru-AI/Integuru — first AI agent building permissionless integrations via reverse-engineered internal APIs (4.8k stars) GitHub · github.com ·