Anvia Docs

@anvia/anthropic adapts the Anthropic SDK into Anvia's completion model interface. It supports streaming, tool use, and the full Anthropic messages API.

Install

pnpm add @anvia/anthropic

The Anthropic SDK (@anthropic-ai/sdk) is a transitive dependency.

Quick Start

import { AgentBuilder } from "@anvia/core";
import { AnthropicClient } from "@anvia/anthropic";

const client = new AnthropicClient({ apiKey });
const model = client.completionModel("claude-opus-4-6");

const agent = new AgentBuilder("support", model)
  .instructions("Answer clearly and concisely.")
  .build();

const response = await agent.prompt("Explain the agent runtime.").send();
console.log(response.output);

Client Configuration

type AnthropicClientOptions = {
  apiKey?: string;     // required unless `client` is provided
  baseUrl?: string;    // custom endpoint
  client?: Anthropic;  // pre-configured Anthropic SDK instance
};

Provider clients only use explicit constructor values. They do not read environment variables.

// Standard
const client = new AnthropicClient({ apiKey });

// Custom base URL (e.g., proxy)
const client = new AnthropicClient({
  apiKey,
  baseUrl: "https://my-proxy.com",
});

// Pre-configured SDK instance
const client = new AnthropicClient({ client: existingAnthropicInstance });

Completion Models

// Default model: claude-sonnet-4-20250514
const model = client.completionModel();

// Specific model
const model = client.completionModel("claude-opus-4-6");

The model implements StreamingCompletionModel:

// Non-streaming
const response = await model.completion(request);

// Streaming
for await (const event of model.streamCompletion(request)) {
  console.log(event);
}

Model Listing

const models = await client.listModels();
console.log(models.data.map((m) => m.id));

Fetches available Anthropic models and returns a normalized ModelList.

Streaming with Agents

Anthropic streaming works transparently with AgentBuilder:

const agent = new AgentBuilder("writer", model)
  .instructions("Write concise technical summaries.")
  .build();

// Stream events
for await (const event of agent.prompt("Summarize the runtime.").stream()) {
  if (event.type === "text_delta") {
    process.stdout.write(event.text);
  }
}

Tool Use

Anthropic models support tool use through the standard Anvia tool system:

import { createTool } from "@anvia/core";
import { z } from "zod";

const search = createTool({
  name: "search_docs",
  description: "Search documentation.",
  input: z.object({ query: z.string() }),
  output: z.object({ results: z.array(z.string()) }),
  async execute({ query }) {
    return { results: [`Result for: ${query}`] };
  },
});

const agent = new AgentBuilder("support", model)
  .tool(search)
  .defaultMaxTurns(3)
  .build();

Error Handling

Constructor throws when neither client nor apiKey is supplied
listModels() throws ModelListingError when the provider request fails
Model operations throw Anthropic SDK errors directly

Models for model name conventions
Anthropic Reference for full API types
Getting Started for a first agent walkthrough