Use with Anthropic

Prerequisites

Python
TypeScript

pip install prompty[jinja2,anthropic]

npm install @prompty/core @prompty/anthropic

You also need an Anthropic API key.

1. Write the `.prompty` File

Create anthropic-chat.prompty:

---
name: anthropic-chat
description: Chat completion with Anthropic Claude
model:
  id: claude-sonnet-4-6
  provider: anthropic
  connection:
    kind: key
    endpoint: ${env:ANTHROPIC_ENDPOINT:https://api.anthropic.com}
    apiKey: ${env:ANTHROPIC_API_KEY}
  options:
    temperature: 0.7
    maxOutputTokens: 1024
inputs:
  - name: topic
    kind: string
    default: the history of computing
template:
  format:
    kind: jinja2
  parser:
    kind: prompty
---
system:
You are a knowledgeable assistant who gives clear, concise explanations.

user:
Tell me about {{topic}}.

2. Set Your API Key

Create a .env file in the same directory:

ANTHROPIC_API_KEY=sk-ant-your-key-here

3. Run It

Python
TypeScript

import prompty

result = prompty.execute(
    "anthropic-chat.prompty",
    inputs={"topic": "quantum computing"}
)
print(result)

import { execute } from "@prompty/core";
import "@prompty/anthropic"; // registers the provider

const result = await execute(
    "anthropic-chat.prompty",
    { topic: "quantum computing" }
);
console.log(result);

Step-by-Step Execution

If you need more control, use the individual pipeline stages:

Python
TypeScript

import prompty

# Load the prompt
agent = prompty.load("anthropic-chat.prompty")

# Render + parse → message list (no LLM call)
messages = prompty.prepare(agent, inputs={"topic": "quantum computing"})
print(messages)  # inspect before sending

# Execute + process → final result
result = prompty.run(agent, messages)
print(result)

import { load, prepare, run } from "@prompty/core";
import "@prompty/anthropic";

const agent = await load("anthropic-chat.prompty");
const messages = await prepare(agent, { topic: "quantum computing" });
console.log(messages); // inspect before sending

const result = await run(agent, messages);
console.log(result);

Available Models

Anthropic’s Claude model family includes:

Model	Best for
`claude-opus-4-0`	Complex reasoning, research, multi-step tasks
`claude-sonnet-4-6`	Balanced performance and speed (recommended default)
`claude-haiku-3-5`	Fast responses, simple tasks, lower cost

Change the model by updating model.id in your .prompty file:

model:
  id: claude-haiku-3-5

Custom Endpoint

If you’re using an Anthropic-compatible proxy or a different base URL, set the endpoint explicitly:

model:
  connection:
    kind: key
    endpoint: https://my-proxy.example.com
    apiKey: ${env:ANTHROPIC_API_KEY}

Or use an environment variable:

model:
  connection:
    kind: key
    endpoint: ${env:ANTHROPIC_ENDPOINT}
    apiKey: ${env:ANTHROPIC_API_KEY}

Key Differences from OpenAI

Aspect	OpenAI	Anthropic
Provider	`provider: openai`	`provider: anthropic`
Default endpoint	`https://api.openai.com/v1`	`https://api.anthropic.com`
Model names	`gpt-4o`, `gpt-4o-mini`	`claude-sonnet-4-6`, `claude-haiku-3-5`
API key prefix	`sk-`	`sk-ant-`
Responses API	✅ Supported	❌ Not supported
Image generation	✅ DALL-E	❌ Not available