Waas | Devpost

WaaS - Workforce as a Service

Inspiration

There’s no deep backstory here—just a strong urge to be part of something big. When I saw Bolt was running the world’s largest hackathon, I knew I wanted in. But the problem? All my ideas either needed a team, were likely to be done by someone else, or would take way too long to build solo. So naturally, I procrastinated... until five days before the deadline.

With the clock ticking, I decided to go solo and build something I could actually finish—an agentic JavaScript framework. No team? No problem. I went full prompt overload on Gemini and used Google AI Studio’s long-context feature to help scaffold the core architecture.

Even then, it wasn’t easy—building agentic frameworks burns through tokens fast, so a lot of my time was spent debugging, refining, and watching those sweet tokens go burn baby burn. But I finally got something working, tested it with demo organizations, and that was my green light to keep going. After that, I used the same prompt madness on Bolt to help me build the UI around the framework.

Honestly, looking back—I’m just proud I made an entry.

What it does

WaaS (Workforce as a Service) is a fully browser-based platform that lets you build, simulate, and observe autonomous agent organizations. You can define hierarchical structures, assign roles, give agents tools, and run goal-based simulations to see how they collaborate.

Each agent operates in an environment, uses tools like writing, searching, analyzing, and even initiating meetings when blocked. SOPs (Standard Operating Procedures) guide repeatable workflows, while the orchestrator ensures everything runs smoothly.

From ad agencies to content studios, WaaS lets you model and observe how AI agents might work as a real team—without any backend or server required.

How we built it

The core framework was built in TypeScript + React, with all agent logic running in the browser using the Gemini API. I used IndexedDB for local persistence and structured the platform around modular components: agents, environments, tools, mail systems, orchestrator, SOPs, and conversations.

The base framework came from my work inside Google AI Studio, where I designed the architecture through intensive prompting. Then I wired up a visual dashboard (the Bolt UI) to interact with and simulate agent teams. It all runs locally. You just need an API key and a browser.

And yes—this is fully client-side agent simulation. No server needed. That alone is why I still stand behind choosing JavaScript over Python. Was it the best choice? Maybe not. But I can justify it.

Challenges we ran into

Tokens. Agentic frameworks eat through them like snacks.
Getting long-context planning to work efficiently inside Google AI Studio while maintaining a coherent architecture.
Debugging roles, tools, and the mail system—all while racing the clock with no team.
Designing a UI that cleanly visualizes a full org of autonomous agents without overcomplicating the flow.

Accomplishments that we're proud of

I built and shipped a working agentic framework and dashboard in 5 days—solo.

The system runs directly in the browser. Agents collaborate, assign tasks, solve problems, and even call meetings. I have prebuilt org demos, a growing toolset, SOPs, and a framework that others can extend.

But above all—I showed up and made an entry.

What we learned

Prompt engineering is an underrated form of solo dev leverage.
JavaScript may not be the obvious choice for agent frameworks—but if you do it right, it opens up cross-platform possibilities: browser, desktop, mobile. No backend required.
Agent-based systems are powerful. If this framework evolves, we could model any organization—real or hypothetical—and solve problems with AI teams.

What’s next for WaaS

I'm going to keep working on WaaS.

Why? Because the potential is wild. If this framework continues to evolve, it could become a plug-and-play base for solving problems across industries—with fully autonomous orgs powered by LLM agents.

So yeah, if you’re a dev and you want to give me a Christmas gift—build something cool on WaaS. I'll take it.