Anthropic Provider

Anthropic Claude models with native tool-use API support. Uses Anthropic's block-based message format internally.

Install

bashpip install "mtpx[anthropic]"

Or install the SDK directly:

bashpip install anthropic

API Key Setup

Option 1: .env file (recommended)

1. Install dotenv support:

bash   pip install python-dotenv

Or with the MTP extra:

bash   pip install "mtpx[dotenv]"

1. Create a .env file in your project root:

text   ANTHROPIC_API_KEY=sk-ant-your_key_here

1. Load it in your code before creating the provider:

python   from mtp import Agent

   Agent.load_dotenv_if_available()  # reads .env file

Get an API key at console.anthropic.com.

Option 2: System environment variable

bash# Linux/macOS
export ANTHROPIC_API_KEY="sk-ant-..."

# Windows PowerShell
$env:ANTHROPIC_API_KEY="sk-ant-..."

Quick Start

pythonfrom mtp import Agent
from mtp.providers import Anthropic

Agent.load_dotenv_if_available()  # loads ANTHROPIC_API_KEY from .env

provider = Anthropic(model="claude-3-5-sonnet-20241022")
tools = Agent.ToolRegistry()
agent = Agent(provider=provider, tools=tools)

reply = agent.run_loop("What is 25 * 4 + 10?")
print(reply)

Parameters

Capabilities

Recommended Models

• claude-3-5-sonnet-20241022 — Best balance of speed and capability (default)
• claude-3-5-haiku-20241022 — Fastest, cheapest
• claude-3-opus-20240229 — Most capable, slowest

Multimodal Support

Anthropic supports images and files (PDFs, documents) natively:

pythonfrom mtp.media import Image, File

provider = Anthropic(model="claude-3-5-sonnet-20241022")
agent = Agent(provider=provider, tools=tools)

# With image
reply = agent.run_loop({
    "content": "Describe this image",
    "images": [Image(filepath="photo.jpg")],
})

# With PDF document
reply = agent.run_loop({
    "content": "Summarize this document",
    "files": [File(filepath="report.pdf")],
})

Full Example

pythonfrom mtp import Agent
from mtp.providers import Anthropic

Agent.load_dotenv_if_available()

provider = Anthropic(
    model="claude-3-5-sonnet-20241022",
    max_tokens=4096,
    temperature=0.0,
)

tools = Agent.ToolRegistry()
agent = Agent(provider=provider, tools=tools, debug_mode=True)

reply = agent.run_loop(
    "Calculate (25 * 4) + 10 and explain the steps.",
    max_rounds=4,
)
print(reply)

Notes

• Anthropic uses a different internal message format (block-based with tool_use and tool_result content types), but MTP normalizes everything to the same event stream and ExecutionPlan semantics.
• System prompts are sent as the system parameter, not as a message role.

Source

src/mtp/providers/anthropic_provider.py