A 20-person team just hit $36M ARR in 45 days with no-code AI agents

Genspark’s Super Agent is a textbook case of building ruthlessly fast and letting product-led growth do the rest. The company pivoted entirely from AI search to autonomous agents in April 2025, and the market didn’t just accept the change — it sprinted toward them. With a lean 20-person crew and zero paid advertising, they reached a $36 million annualized run rate in 45 days. That’s the kind of traction that forces you to pay attention, especially when it comes entirely from users sharing the product on their own.

The platform is a no-code playground built on a multi-model backbone. Under the hood, GPT‑4.1 crunches through documents with a 1M-token context window and spits out the strict JSON required to keep the system’s 80-plus integrated tools from tripping over each other. For visual tasks, they lean on the GPT‑image‑1 model. CEO Eric Zhu frames the OpenAI relationship as more than a vendor deal: “We chose OpenAI not just for model performance across modalities, but for developer experience.” Regular sit-downs with OpenAI solutions architects helped the team fine-tune workflows and avoid the scaling bottlenecks that usually plague small teams shipping this fast.

The feature generating the most noise is Call For Me. It sidesteps clunky chatbot scripts by pairing the OpenAI Realtime API for live conversation with a second “shadow” model that monitors the interaction through a message queue. That dual-layer approach lets it handle the messy reality of phone trees, hold music, and confused receptionists without falling apart. A use case in Japan went viral for a reason most product managers wouldn’t put on a roadmap: users started deploying the agent to make resignation calls to their employers. It’s a deeply uncomfortable, very human interaction, and that people trusted an AI to handle it says something about the system’s conversational fluency.

Eight major features shipped in 70 days. That velocity from a team this size is absurd, and it only works because the platform abstracts away all the orchestration complexity. Users don’t configure anything. They describe an outcome — “make me a vaporwave pitch deck” — and the agent drafts slides, generates stylized images, and compiles the final file. It’s easy to dismiss this as another AI wrapper until you look at the revenue curve. Genspark is now expanding into new verticals, and if the organic growth pattern holds, the question isn’t whether they can scale but how fast they’ll redefine what a small team can build.