v0.4.2 — sites, apps & agents on one runtime

Ship AI to production.
Not to a sandbox.

Ship4AI is the deployment platform for AI-native software — sites, apps, and agents on one runtime. Write TypeScript, push to a branch, get a live URL with logs, eval, and rollback. No canvases. No nodes. No nonsense.

Start shipping

npx create-ship4ai

47ms

Cold start

12 regions

Edge deploys

$0.0001

Per invocation

apps/docs-copilot.ts

1import { app } from "ship4ai"
2import { chat, rag, embed } from "ship4ai/ai"
3
4export default app({
5  name: "docs-copilot",
6  model: "claude-opus-4.6",
7  routes: {
8    "/": site({ content: "./content" }),
9    "/api/chat": chat({ retriever: rag("./docs") }),
10  },
11  domain: "docs.acme.dev",
12})

live execution

iad1 · 218ms

→Email received from elena@acme.dev

→Classified: bug_report · confidence 0.94

→Tool call createTicket()

✓LIN-2847 created · routed to @core-team

✓Slack #support notified

Done· 1.2s · 4 tool calls · $0.0034

Sites, apps & agents on one runtime12.4M invocations / 24hp99 latency 218ms0 cold starts on edge runtimeTypeScript-first — no canvas, no YAML$0.0001 / invocationStreaming on by defaultEval suite runs on every pushSites, apps & agents on one runtime12.4M invocations / 24hp99 latency 218ms0 cold starts on edge runtimeTypeScript-first — no canvas, no YAML$0.0001 / invocationStreaming on by defaultEval suite runs on every push

02 — Showcase

Sites. Apps. Agents. One runtime.

Three shapes of AI-native software, copy-paste ready. Every example is a working project, deployed on the same runtime you'll use.

Production-tested

Every example here has been running in customer environments for months. Fork the repo, swap the keys, ship it.

agents/site.ts

1import { site } from "ship4ai"
2import { semanticSearch, summarize } from "ship4ai/ai"
3
4export default site({
5  name: "acme-docs",
6  content: "./content",
7  features: {
8    search: semanticSearch({ model: "openai/text-embed-3" }),
9    summary: summarize({ model: "claude-opus-4.6" }),
10  },
11  domain: "docs.acme.dev",
12})

live execution

iad1 · 218ms

→Build · 142 pages · 3,204 chunks

→Embedded in 4.8s

✓Deployed · docs.acme.dev

✓Search p50 34ms

03 — Primitives

The runtime is the product.

We are not another LLM wrapper. Ship4AI gives you the eight primitives every production AI site, app, or agent eventually needs — built into the platform, not bolted on.

Preview deploys

Every git push gets a unique URL. Test your site, app, or agent in isolation before merging to main.

$git push → preview.ship4.ai/abc123

Observability built in

Every tool call, every token, every dollar — traced and searchable.

$ship4ai logs --tail

Secrets, scoped

Bring your own keys. Scope per env. Never logged.

$ship4ai env add OPENAI_KEY

Atomic rollbacks

Bad deploy? One command, gone. Previous version is one second away.

$ship4ai rollback

Eval on every push

Define golden cases once. Block deploys when scores regress.

$ship4ai eval --threshold 0.85

Edge runtime

Cold starts under 50ms in 12 regions. Streaming on by default.

$runtime: 'edge'

Trigger on anything

Webhooks, cron, email, queues. Wire your agent to the real world in two lines.

$on: { cron: '0 9 * * *' }

Stateful by default

Memory, sessions, embeddings — first-class across sites, apps, and agents. No vector DB to wire up.

$ctx.memory.recall(query)

04 — Workflow

Four steps. No meeting required.

The same workflow you already use for web apps — extended to AI. If you can ship a Next.js project, you can ship a production agent.

Write

TypeScript. Your editor, your stack. Pick site, app, or agent — the shape of the export is the only thing that changes.

import { site, app, agent }

from 'ship4ai'

export default app({...})

Push

Git push to any branch. Preview URL in your PR within seconds.

$ git push origin feat/triage

→ deploying preview...

Eval

Golden cases run on every push. Regress and the deploy is blocked.

$ ship4ai eval

✓ 47/48 passed · 0.91 avg

Ship

Promote to prod with one command. Atomic. Reversible. Logged.

$ ship4ai promote

→ live at agent.acme.com

Works with GitHub, GitLab, Bitbucket

CLI, dashboard, or REST — pick your poison

SOC 2 Type II, HIPAA-ready

Built for engineers shipping AI to production

12.4M

Invocations / 24h

218ms

p99 latency

47ms

Cold start

Edge regions

99.99%

Runtime uptime

DSLs to learn

SOC 2

Type II audited

First-class

"The runtime is the product. We give you the primitives every AI site, app, and agent needs — not another LLM wrapper pretending to be infrastructure."

Engineering

Principle 01

"Eval on every push. If golden cases regress, the deploy is blocked. That is the only way we trust shipping AI on Friday afternoons."

Reliability

Principle 02

"Vercel showed the world what the frontend runtime could be. We are doing the same thing for AI. Push code, get a URL, watch it run."

Product

Principle 03

05 — Pricing

Pay for invocations. Not seats.

Your team can grow without your bill exploding. We charge for the work the agents do, not the people who deploy them.

Hobby

$0/forever

For tinkering, weekend projects, the AI app that scratches your itch.

Start free

10k invocations / month
1 production project
Community support
Edge runtime

Pro

$24/month

For builders shipping real products. Everything you need, no enterprise nonsense.

Start shipping

1M invocations / month
Unlimited sites, apps & agents
Eval suites + previews
Custom domains
Priority support

Scale

Custom

For teams shipping agents at the edge of what is possible. Talk to us.

Talk to sales

Unlimited everything
Dedicated runtime
SOC 2, HIPAA, BAA
Bring your own cloud
Slack channel + SLA

One runtime for AI sites, apps & agents

Stop demoing.
Start shipping.

Your competitors are still building flowcharts. You can have a real AI site, app, or agent running in production before lunch.

Deploy your first project

$npx create-ship4ai my-app

Ship AI to production.Not to a sandbox.