Help build the agentic web.
Steel funds the research and benchmarks that show what agents can really do.
What fits
- Papers on web agents
- Public benchmarks anyone can rerun
- Evals on real websites
FAQ
You get $1,000 when you start and another $500 when your paper is published. You can renew when you start your next paper.
There's no waitlist. The program is for researchers at reputable institutions, PhD students, faculty, and teams who maintain public benchmarks.
Any research that puts browsers in the method: agent experiments, evals and benchmarks, or web data collection. Credits cover all of it.
Your account moves to a standard plan when the credits run out. Publish and the grant renews for your next paper.
Building a startup?
Get up to $25,000 in free plan access, usage credits and help from the Steel team.
Steel for startups