Anthropic

@anvia/anthropic runs Anvia completions, agents, extractors, and pipelines on Claude. It adapts Anthropic's Messages API into the provider-neutral completion contract used throughout the SDK.

Use the same adapter with either Anthropic's direct API or Claude on Google Vertex AI. Only the client setup changes; the model passed to the rest of the application remains an Anvia completion model.

Choose a connection

Deployment	Client	Authentication
Anthropic API	AnthropicClient	ANTHROPIC_API_KEY
Claude on Vertex AI	AnthropicVertexClient	Google Application Default Credentials
Anthropic-compatible endpoint	AnthropicClient with baseUrl	Endpoint-specific API key

Start with the direct API unless the application already runs Claude through Google Cloud or needs an Anthropic-compatible gateway.

Create a model

import { AnthropicClient } from '@anvia/anthropic'

const anthropic = new AnthropicClient({
  apiKey: process.env.ANTHROPIC_API_KEY,
})

export const model = anthropic.completionModel(
  'claude-sonnet-4-20250514',
)

Pass model to an agent, direct completion, extractor, or pipeline. Keep the client and credentials in server-only code.

What the adapter supports

The completion adapter supports streaming, tools, tool choice, image input, PDF input, and Anthropic reasoning content. It does not declare support for Anvia final output schemas.

That distinction matters: use tool-backed extraction when you need structured data, and do not assume createParsedCompletion(...) or agent .outputSchema(...) is portable to this provider.

In this section

• Setup installs the package and creates a client.
• Capabilities covers streaming, tools, media, reasoning, and structured data.
• Model options explains model IDs and request settings.
• Vertex AI configures Google authentication.
• Production covers boundaries, testing, and operational checks.