The General Intelligence Company is on a mission to enable the one-person, one billion-dollar company. Their product, Cofounder, lets anyone automate work with natural language by plugging into existing tools, writing automations, and orchestrating workflows through specialized agents. And they mean it: their entire engineering operation runs on Cofounder itself, with agents writing code, creating pull requests, and testing features around the clock.
The problem: local dev doesn't scale for agents
To build Cofounder, the General Intelligence Company needed agents that could log in, navigate, and test new features the way a real user would, across dozens of preview environments at once. Every run needed to be recorded so engineers could review what happened without pulling a branch or spinning up a local environment. And it all had to scale without the team managing browser infrastructure themselves.
Setting up local browser environments for each agent wasn't an option. It's slow to configure, brittle to maintain, and impossible to parallelize cleanly.
"If you're building a web app, it's just significantly easier to use something like Browserbase than set up a full browser internally." — Andrew Pignanelli, CEO
The solution: Browserbase as the eyes of every agent
The General Intelligence Company needed agents that could log in, navigate, and test new features the way a real user would, across dozens of preview environments at once. To make that work, they connected Browserbase to their Vercel preview environments and brought in Agent Identity to handle authentication, so agents could access any environment without manual credential management. Every time an agent creates a pull request, a new preview environment spins up, gets paired with a Browserbase session, and the agent logs in, tests the feature, and records the entire session from start to finish.
Here's how the full loop works:
- An agent writes and submits a PR. It gets deployed to a Vercel preview environment automatically.
- Browserbase spins up a browser session for that preview environment.
- A browser agent tests the feature, navigating the app the way a real user would, using both vision and DOM to understand the interface.
- A video replay is generated and attached to the PR for human review.
- If the test passes, the agent merges the PR.
The result is a fully autonomous code-to-merge pipeline, where engineers stay unblocked and PRs ship without manual QA becoming a bottleneck.
To power the agent's interactions with the web, General Intelligence Company uses Stagehand, Browserbase's AI SDK for browser agents. Stagehand gives the agent a natural way to interact with any web interface, combining visual reasoning with DOM understanding to handle real-world complexity. Every session is captured through Browserbase's Observability, giving engineers rich logs, live view, and session replay without ever pulling a branch.
"Browserbase is the easiest way to do that. It's also the cheapest way to do that." — Andrew Pignanelli, CEO
The results: a 5-person team shipping like a company of thousands
- 4 to 5x faster engineering compared to a traditional development workflow
- ~50 PRs merged per day with a team of 5 engineers
- 10 to 20 concurrent browser sessions running at any given time
Browserbase handles the browser infrastructure so their engineers can stay focused on building.
"We have an engineering team of five and we ship a quarter of the amount that Stripe does, and they have 10,000 people." — Andrew Pignanelli, CEO
Enabling the one-person, one billion-dollar company
The General Intelligence Company envisions a world where anyone with an idea can start a business and have a full staff of agents running it around the clock. That future requires agents that can navigate and take action across the web, just like a human would, and Browserbase is the infrastructure that makes that possible.
Ready to build? Join companies like the General Intelligence Company who are building the future of autonomous engineering with browser agents. Give your agents the web →
