We're the full-stack, agent-native platform for AI agents. Bring the framework you already use, and every app gets runtime services (e.g., storage, sandboxes, an AI gateway) built in, not bolted on.

From local government to indie developers, people are already building with us.

npm create agentuity -- --name my-app --framework nextjs # or nuxt, sveltekit, astro, hono
cd my-app
npx agentuity dev # local dev; add --public for a shareable HTTPS URL
npm run deploy # buildpack deploy to the cloud
# Works with npm, pnpm, yarn, or bun!
# Prefer a global CLI? curl -fsSL https://agentuity.sh | sh

The whole app is a typed AI endpoint you can drop into any framework. The AI Gateway gives you one credential for every provider, with unified billing, and switching models is a simple string change.

1. The AI call (src/translate.ts)

import { AIGatewayClient, getAIGatewayCompletionText } from '@agentuity/aigateway';
import { z } from 'zod';
const Input = z.object({
 text: z.string(),
 toLanguage: z.enum(['Spanish', 'French', 'German']),
});
// One credential covers every provider. The model below is just a string:
// swap it for any other, e.g. 'anthropic/claude-opus-4-8' or 'googleai/gemini-3.5-flash'.
const ai = new AIGatewayClient();
export async function translate(input: z.infer<typeof Input>) {
 const { text, toLanguage } = Input.parse(input);
 const completion = await ai.complete({
 model: 'openai/gpt-5.5',
 messages: [{ role: 'user', content: `Translate to ${toLanguage}:\n\n${text}` }],
 });
 return { translation: getAIGatewayCompletionText(completion) };
}

2. The route, with services built in (src/index.ts)

import { serve } from '@hono/node-server';
import { Hono } from 'hono';
import { agentuity } from '@agentuity/hono';
import { translate } from './translate';
const app = new Hono();
app.use('*', agentuity()); // injects c.var.kv, c.var.vector, c.var.logger, c.var.sandbox, and more
app.post('/api/translate', async (c) => {
 const body = await c.req.json();
 const cacheKey = `${body.toLanguage}:${body.text}`;
 // Key-value storage comes ready through the middleware, provisioned and connected for you
 const cached = await c.var.kv.get<{ translation: string }>('translations', cacheKey);
 if (cached.exists) return c.json(cached.data);
 const result = await translate(body);
 await c.var.kv.set('translations', cacheKey, result, { ttl: 3600 });
 return c.json(result);
});
serve({ fetch: app.fetch, port: Number(process.env.PORT ?? 3000) });

Prefer Next.js, SvelteKit, or Astro? The same services work there. Import a client like @agentuity/keyvalue or @agentuity/vector directly, in a route, a script, or a cron job.

Everything you need to build and ship AI agents:

Build

• TypeScript-first: typed service clients and schemas, autocomplete everywhere, runs on Bun or Node 24+
• Bring any framework: Next.js, Nuxt, SvelteKit, Astro, Hono, or your own. agentuity deploy auto-detects and builds it
• Standalone service clients: import @agentuity/keyvalue, @agentuity/vector, @agentuity/queue, and the rest into any app or script; they work outside Agentuity too
• Frontend included: ship your framework's UI alongside your API on a single deploy
• Bring your own stack: pair with the OpenAI Agents SDK, Mastra, and LangChain, or the AI SDK; validate with any Standard Schema-compatible library, like Zod, Valibot, or ArkType
• Agent-friendly CLI: --json output, --explain previews, --dry-run validation, --services to scaffold storage on create

Connect

• AI Gateway : one key for OpenAI, Anthropic, Google AI, Groq, and more
• Storage and services: key-value, vector search, object storage, durable streams, queues, schedules, tasks, email, and webhooks, plus Postgres via DATABASE_URL
• Sandboxes : isolated Linux containers with language runtimes (e.g., Bun, Node, Python) and coding agents (e.g., opencode); pause and resume with checkpoints, or snapshot a sandbox to reuse later
• Custom domains: declare them in agentuity.json; DNS is validated on deploy and TLS is provisioned automatically on first request
• Deploy: managed cloud with GitHub auto-deploy and PR preview environments; bring-your-own-cloud (VPC, on-prem, edge) via the Gravity Network

Monitor

• Observability : OpenTelemetry traces and structured logs through @agentuity/telemetry (or your own OTel collector); the Hono middleware injects c.var.logger, c.var.tracer, and c.var.meter
• Sessions and logs: session records, timelines, and deployment logs to inspect in the Web App
• Container access: SSH and SCP into running deployments with agentuity cloud ssh

v3 trades createAgent() for plain framework code and standalone service clients. Run npx @agentuity/migrate --v2-to-v3 to convert an existing project, or read the migration guide. Staying on v2 for now? The v2 docs still cover the v2 SDK in full.

• SDK: runtime, CLI, service clients, framework integrations
• Examples: training, integrations, features
• Docs: guides and reference (v3/latest)
• v2 Docs: reference for projects still on v2
• Web App: manage projects, services, billing, and more
• Website: product, customers, pricing, and our blog
• Discord: join 250+ developers to chat with us and other builders, or ask for help

Our SDK is open source, so contributions are welcome! See the repo for guidelines.