describe an agent. get a working one.
features
build & test ↳ describe an agent in plain english ↳ generates real, working eve files ↳ tested against a live eve runtime before you see it ↳ generation and testing run as one durable step, survives crashes ↳ if a build fails, the reason is shown and you can retry with one click
chat & share ↳ inspect every file with syntax highlighting ↳ export the full agent as a zip ↳ share a live link to any agent you build ↳ connect to your agent right after it's built, no install needed ↳ chat with it live, with markdown-formatted replies
reliability ↳ reload the page anytime, your agent and chat pick up right where you left off ↳ idle or closed sandboxes shut down automatically, nothing left running ↳ warned before disconnect, never cut off without notice ↳ your history is private, delete any entry or clear it all
built with ▲ vercel products ↳ next.js: the app itself ↳ ai gateway: routes the generation request to a model ↳ ai sdk: streams the model's response ↳ sandbox: tests every agent against a real eve runtime, then runs it live so you can talk to it ↳ workflow sdk: runs generate and test as one durable step, survives crashes ↳ blob: stores each generated agent and its live chat session, so shared links and reloads stay in sync ↳ cron: sweeps stale sandbox sessions on a schedule ↳ firewall: rate limits generation and connect requests to keep usage fair ↳ ai elements: the chat interface, task progress ui, shimmer loading text ↳ streamdown: renders code and markdown cleanly ↳ shadcn/ui: every ui component ↳ vercel: hosts and deploys the app ↳ analytics: tracks real usage without slowing anything down
open source: https://github.com/AbhiVarde/tryeve
live: https://tryeve.abhivarde.in/
would love feedback from the eve or vercel team, open to adding more free vercel products or useful features.
x: https://x.com/varde_abhi/status/2079565257526354255 linkedin: https://www.linkedin.com/feed/update/urn:li:activity:7485331089419579393/


