Integuru

Integuru (YC W24) is the purest expression of this category's core idea and the cheapest way to test whether it works for you. You capture a HAR file and cookies from a session, point the agent at it, and an LLM identifies the request chain and data dependencies, builds a dependency graph, and emits runnable Python that calls the platform's internal API directly. The core agent is AGPL-3.0, fully open source, and has drawn 4,700+ GitHub stars.

The generated integrations are legitimately fast. The company reports roughly three-second completions with 99.9%+ success rates, and says the approach handles over a million API calls per month per platform, with users building low-latency APIs for platforms like Robinhood and trucking TMS systems.

What you do not get is everything after generation. There is no observability layer watching for the 3am redesign, no auto-repair, no MFA or session-rotation handling as a managed service, no support, no SLA and no one to call. The documentation is essentially a README. AGPL-3.0 is also a real constraint for commercial products. Integuru scores near the top of the category on accessibility because it costs nothing, and near the bottom on reliability because the entire maintenance burden, which is where these projects actually die, transfers to you on day one.