Define services, databases, queues, and workflows in TypeScript. Vy deploys them to the edge with end-to-end tracing and a typed infrastructure graph humans and AI agents can both reason about.
7-day free trial · Usage-based pricing · $15/mo minimum
1// app/orders.ts2import { service } from "vy/service";3import { database } from "vy/storage";4import { topic } from "vy/pubsub";56export const db = database("orders");7export const events = topic("order.created");89export default service();84% of developers report using or planning to use AI coding tools. 46% don't trust the output, up from 31% a year ago. The bottleneck isn't writing code anymore, it's understanding, debugging and operating what gets generated.
Generation is fast. Production is the gap.
AI app builders and coding agents can scaffold a full-stack prototype in an afternoon. Getting that prototype deployed, traced, and safe to iterate on is still the slow part, and it's where most generated apps stall.
Runtime truth beats reading code.
When AI wrote half the file last week, the only honest source is what actually happened in production. Logs, grep and intuition don't scale, you need a runtime that tells you what your system is really doing.
Six vendors deep, still no full picture.
Frontend host, backend host, database, queue, storage, APM, secrets, CI. Each owns a slice of the request, none of them see the whole one. Most of the work is wiring, not product.
Vy unifies what most teams stitch from six vendors: typed services co-located with your frontend, a runtime graph humans and AI agents can introspect, and tracing that follows a request from the browser to the database and back.
Frontend and backend, one typed deploy.
Put your frontend and backend in the same repo. Vy deploys both, generates a typed client across the boundary, and provisions the database, queues, cron, and storage your code declares, no Terraform, no glue SDKs, no secrets dashboard to babysit.
1// apps/web/app/orders/page.tsx2import { backend } from "vy/clients";34export default async function Page() {5 const orders = await backend.orders.list(); // fully typed6 return <OrderList data={orders} />;7}89// apps/backend/orders.service.ts10export const list = api.get("/orders", () => db.orders.findMany());Tracing across the entire stack.
Follow a single request from the browser click, through the API route, into the queue, to the database query and back, in one timeline. Built into the runtime, no separate APM to configure. The runtime graph is the source of truth that humans and AI agents both read from.
A live environment for every branch.
Each pull request, yours or your agent's, gets its own URL, its own database branch, and its own trace view. Pay-per-request billing means an idle preview costs nothing, so reviewing turns into clicking real behavior instead of staring at diffs.
Vy gives you the building blocks for the products teams actually ship today, from AI agents to SaaS backends, without stitching together a dozen vendors.
AI apps & agents
Ship chatbots, copilots, and autonomous agents with streaming responses, tool calls, and durable memory, without wiring five services together.
SaaS backends
Multi-tenant APIs with auth, billing, and a Postgres database. Type-safe end to end, deployed globally on the first push.
Stateful actors
Single-instance objects with durable state, sessions, rooms, agent memory. Vy handles persistence and lifecycle.
Background jobs & pipelines
Long-running work that survives restarts: payment flows, ETL, onboarding sequences. Resumes exactly where it left off.
Internal tools & dashboards
Stand up admin panels, ops consoles, and reporting on top of your real data, no separate analytics stack.
Event-driven products
React to webhooks, user events, or upstream changes. Fan out work across services without managing a broker.
Media & file workflows
Uploads, transformations, and signed downloads. Browser uploads go straight to storage, then trigger your code.
Scheduled & recurring work
Daily reports, reminders, sync jobs, cleanup tasks. Schedule it once, run it forever, see every execution.
APIs for AI agents
Expose your product to LLMs over MCP and HTTP. Agents get typed tools, traces, and preview environments to test against.
A Vercel-style developer experience for the whole stack, frontend, backend, database and jobs, deployed to a global edge network. Requests run close to your users, idle previews and dormant endpoints don't show up on the bill.
You pay whether traffic arrives or not. Scale-to-zero is rarely the default.
Pay when a request runs. Spin up dozens of services and previews without watching the meter.
From AI prototype to production.
Define your stack in TypeScript. Vy deploys frontend and backend together, traces every request end-to-end, and gives every branch its own preview environment, on a typed graph humans and AI can both reason about.