Anthropic Claude

Use Anthropic Claude models with PraisonAI Agents

Environment Variables

Setting MODEL_NAME=anthropic/claude-3-5-sonnet is enough — no need to set OPENAI_BASE_URL. The ANTHROPIC_API_KEY environment variable is consulted; OPENAI_API_KEY is not used as a fallback.

If you just want the Anthropic default, setting ANTHROPIC_API_KEY alone is enough — no MODEL_NAME, no llm=. Agent(...) picks anthropic/claude-3-5-sonnet-latest and routes it through litellm. See Provider Auto-Detection.

import os
from praisonaiagents import Agent

os.environ["MODEL_NAME"] = "anthropic/claude-3-5-sonnet"
os.environ["ANTHROPIC_API_KEY"] = "your-anthropic-key"

agent = Agent(name="Claude", instructions="You are Claude")

Models

• Recommended: claude-3-5-sonnet-20241022 (best balance)
• Fast/Cheap: claude-3-5-haiku-20241022 (fastest, affordable)
• Advanced: claude-3-opus-20240229 (highest capability)
• Latest: claude-sonnet-4-5, claude-haiku-4-5 (newest)

Routed model formats

Claude reached through Bedrock, Vertex AI, or OpenRouter is now detected as Anthropic — not misrouted to the OpenAI adapter — so async streaming works for these prefixes too.

Routed model id	Detected provider	Async streaming
bedrock/anthropic.claude-3-5-sonnet-v1:0	Anthropic	✅ supported
vertex_ai/claude-3-5-sonnet@20240620	Anthropic	✅ supported
openrouter/anthropic/claude-3.5-sonnet	Anthropic	✅ supported

See Routed model provider detection for the full detection rules.

Python

# export ANTHROPIC_API_KEY=your-api-key
from praisonaiagents import Agent

agent = Agent(
    instructions="You are a helpful assistant",
    llm="claude-3-5-sonnet-20241022"
)
agent.start("Explain quantum computing simply")

With Tools

# export ANTHROPIC_API_KEY=your-api-key
from praisonaiagents import Agent

def calculate(expression: str) -> str:
    """Evaluate a math expression."""
    return str(eval(expression))

agent = Agent(
    instructions="You are a math assistant",
    llm="claude-3-5-haiku-20241022",
    tools=[calculate]
)
agent.start("What is 25 * 17 + 89?")

Multi-Agent

# export ANTHROPIC_API_KEY=your-api-key
from praisonaiagents import Agent, Task, AgentTeam

analyst = Agent(
    instructions="You analyze data and find insights",
    llm="claude-3-5-sonnet-20241022"
)
reporter = Agent(
    instructions="You write clear reports",
    llm="claude-3-5-haiku-20241022"
)

task1 = Task(description="Analyze market trends", agent=analyst)
task2 = Task(description="Write executive summary", agent=reporter)

agents = AgentTeam(agents=[analyst, reporter], tasks=[task1, task2])
agents.start()

CLI

export ANTHROPIC_API_KEY=your-api-key

# Basic prompt
python -m praisonai "Explain AI" --llm claude-3-5-sonnet-20241022

# With temperature
python -m praisonai "Write a story" --llm claude-3-5-sonnet-20241022 --temperature 0.8

# Run agents.yaml
python -m praisonai

YAML

framework: praisonai
topic: Analyze business strategy
agents:
  analyst:
    role: Business Analyst
    goal: Analyze market opportunities
    instructions: You are an expert business analyst
    llm:
      model: claude-3-5-sonnet-20241022
    tasks:
      analysis_task:
        description: Analyze current market trends and opportunities
        expected_output: Detailed market analysis report

  strategist:
    role: Strategy Consultant
    goal: Develop actionable recommendations
    instructions: You create strategic business plans
    llm:
      model: claude-3-5-haiku-20241022
    tasks:
      strategy_task:
        description: Create strategic recommendations based on analysis
        expected_output: Strategic action plan