---
id: agent-framework-anthropic
version: "1.0.0b260730"
license: unclear
license_treatment: permissive
maintenance: active
---

# agent-framework-anthropic — Anthropic integration for Microsoft Agent Framework.

License: permissive · Maintenance: active · Downloads: 543.2K/mo

## What it is and what it does

This package bridges Microsoft's Agent Framework with Anthropic's API, letting you build agent applications that call Anthropic models. It provides a unified client interface (AnthropicClient) for direct Anthropic endpoints and three hosted transport wrappers—AnthropicFoundryClient, AnthropicBedrockClient, and AnthropicVertexClient—that route requests through Microsoft Foundry, Amazon Bedrock, and Google Vertex AI respectively.

The package supports both streaming and non-streaming responses, and includes structured system block support for Anthropic's prompt caching feature with cache control metadata. It depends on agent-framework-core for the framework scaffolding and anthropic for API communication. The package is in beta and requires Python 3.10 or later.

Use it for:

- Build multi-provider agent applications that can switch between direct Anthropic and hosted transports without changing core logic.
- Implement agents with prompt caching to reduce latency and costs on repeated or similar requests to Anthropic.
- Route agent workloads through Amazon Bedrock or Google Vertex AI for compliance, latency, or organizational requirements.
- Stream agent responses for real-time interaction while maintaining structured system prompts with cache control.
- Integrate Anthropic models into existing Microsoft Agent Framework deployments.

## Worth the install?

AI-flagged interpretation of the facts above — verify before relying.

Connects Agent Framework applications to Anthropic's API, providing client implementations for direct Anthropic endpoints and hosted transports via Microsoft Foundry, Amazon Bedrock, and Google Vertex AI.

Yes, if you are building agents with Microsoft Agent Framework and want to use Anthropic models. The package is actively maintained, has low install friction, carries a permissive MIT license, and offers both direct and multi-cloud transport options. Be aware it is in beta—test thoroughly before production use.

## Install

pip install agent-framework-anthropic
uv add agent-framework-anthropic
poetry add agent-framework-anthropic

## Installing agent-framework-anthropic

Before you install: Low friction install with only two runtime dependencies. Active maintenance—released 15 days ago with recent commits and a healthy repository signal.

License in practice: MIT license (permissive) means you can use, modify, and distribute this package freely with minimal restrictions.

Quickstart:

pip install agent-framework-anthropic --pre

from agent_framework_anthropic import AnthropicClient

client = AnthropicClient()
response = await client.get_response("Hello")

Requires Python 3.10 or later; package is in beta (pre-release) and depends on agent-framework-core and anthropic packages.

Verify before relying:

- Whether streaming and non-streaming response modes have feature parity or known differences.
- Whether prompt caching with ephemeral TTL is supported across all three hosted transports (Foundry, Bedrock, Vertex).
- Performance characteristics and latency overhead of the transport wrappers compared to direct Anthropic API calls.

## Package facts

- License: not declared (permissive)
- Python support: supports_current
- Install friction: low
- Maintenance: active
- Downloads: 543.2K/month (top 15,000 on PyPI)
- Known vulnerabilities: none known

## Tags

anthropic agent framework integration, agent framework anthropic client, bedrock vertex foundry anthropic, anthropic api agent wrapper, structured prompts anthropic agents, multi-provider anthropic transport, agent-framework, anthropic-integration, multi-cloud-transport

[View on SkillFed](https://skillfed.io/packages/agent-framework-anthropic) · [View on PyPI](https://pypi.org/project/agent-framework-anthropic/)