GitHubRepo
Hey everyone.
I've been teaching myself AI agent development for about six months now, and today I'm sharing my first open-source tool. Nervous but excited.
The problem I ran into: I wanted to use Google's Deep Research Agent inside a CrewAI pipeline, but Deep Research runs on the Interactions API with async background polling. CrewAI tools are synchronous. I couldn't find an existing solution that just worked as a native BaseTool, so I built one.
What it does: starts the background research, polls until it's done, enforces a hard timeout so nothing runs forever, and limits how many times an agent can call it per task (because at $2+ per call, that matters). One file, you import it, give it to an agent, done.
It's already updated for the June 2026 Breaking Change (outputs to steps), so it should keep working when Google makes the switch.
I'm still learning, so if you see something I could do better, whether it's the code, the architecture, or just the README, I genuinely want to hear it. This is how I get better.
GitHub:
https://github.com/Nick-is-building/crewai-gemini-deep-research
Thanks for taking a look.
Top comments (0)