Install and setup

Install Anvia's core runtime and the provider package your application needs. This guide uses OpenAI, but the runtime stays provider-neutral.

Before you start

You need an ESM TypeScript project, pnpm, and an API key for your chosen model provider.

Install the packages

Start with the smallest useful stack: the core runtime and one provider.

pnpm add @anvia/core @anvia/openai

Add other packages only when the application needs them.

Package	Use it for
@anvia/server	Exposing agent runs and streams over HTTP.
@anvia/react	Consuming agent state and streams in React.
@anvia/logger	Structured runtime logging.
@anvia/sandbox	Running code in an isolated sandbox.
@anvia/studio	Inspecting agents during development.
zod	Defining schemas for tools and structured output.

Configure your API key

Set the provider key in the server environment. Keep it out of browser code and source control.

OPENAI_API_KEY=your_api_key

Create a model

Provider clients create models that can be passed to Anvia's runtime primitives.

import { OpenAIClient } from '@anvia/openai'

const client = new OpenAIClient({
  apiKey: process.env.OPENAI_API_KEY,
})

export const model = client.completionModel('gpt-5')

Verify the setup

Run one direct completion before adding an agent. This confirms that the package, credential, and provider model are wired correctly.

import { createCompletion } from '@anvia/core'
import { model } from './model'

const result = await createCompletion(model, {
  input: 'Reply with: Anvia is ready.',
})

console.log(result.text)

If the request returns a response, the setup is ready. Continue with Your first agent.