IDE for AI Agents on Vercel: How Superset Did It
Software development with AI started as a single engineer chatting with a single agent about a local repo. Today, developers direct fleets of agents in the cloud. Yet traditional tools, IDEs, terminals, review systems, were built for the old workflow: one developer, one ticket at a time.
That’s exactly the problem Superset set out to fix. Co-founders Kiet Ho, Satya Patel, and Avi Peltz, all former CTOs at YC-backed companies, built Superset as the IDE for multi-agent development. It runs up to 10 coding agents in parallel, each in its own isolated workspace.
To deliver that kind of parallelism reliably, the infrastructure underneath must match the pace. Without instant provisioning, parallel agents stop being parallel. CI pipelines must be configured per branch, preview environments must be managed by hand, and deploys back up behind one another. For a team running a dozen agents at once, that serialization breaks the product entirely.
Vercel solved that bottleneck. All three founders had built on Vercel at previous companies, so it was the default choice from day one. Superset ran six Next.js projects on Vercel from the start, the web app, marketing site, docs, and three supporting services, and skipped platform engineering entirely.
The results speak for themselves. Every branch a Superset developer or agent creates becomes a preview deployment automatically, often spinning up multiple services. At peak, Superset generates roughly 600 preview deployments a day internally. Every branch gets a live URL, and the team never waits on a deploy queue.
The stack grew alongside the product. Vercel Blob stores agent artifacts. Cron Jobs prevent parallel environments from piling up. BotID filters bots during traffic spikes. Meanwhile, agent orchestration runs on the AI SDK and AI Elements, with AI Gateway handling model routing. Fluid compute carries the scaling load underneath, absorbing parallel tasks as agents fan out.
Resilience matters too. During a Hacker News “Show HN” launch, user counts tripled overnight. Superset absorbed the spike without anyone provisioning infrastructure mid-flight. If a customer reports a bug, agents can spin up, write a fix, generate a preview, and merge code in under thirty minutes. If the fix goes wrong, rollbacks are instant.
The proof is in the team’s own workflow. The team’s setup runs up to a dozen instances at once. Multiple efforts move forward without anyone waiting on serial decisions. Compared to previous dev workflows, Superset’s commit graph looks exponential.
Build time averages around 30 seconds. Deployments run between 1,000 and 1,400 per week. Weekly DAU growth sits between 57% and 64%. For a team of three building the IDE for AI agents on Vercel, those numbers say everything.
