Agents
Agent Runtimes provides a flexible agent architecture built on top of Pydantic AI.
Why Pydantic AI?
We chose Pydantic AI as our foundation because it provides:
- Type Safety — Full type checking with Pydantic models for inputs and outputs
- Structured Outputs — Reliable JSON responses from LLMs with validation
- Tool Calling — First-class support for function tools and MCP toolsets
- Multi-Model Support — Works with Anthropic, OpenAI, Google, Azure, and more
- Production Ready — Well-tested and actively maintained
Creating an Agent
from pydantic_ai import Agent
from agent_runtimes.mcp import get_mcp_toolsets
# Get pre-loaded MCP toolsets
mcp_toolsets = get_mcp_toolsets()
# Create an agent with MCP tools
agent = Agent(
"anthropic:claude-sonnet-4-20250514",
system_prompt="You are a helpful assistant with access to web search.",
mcp_servers=mcp_toolsets,
)
Model Providers
Agents can use models from multiple providers:
| Provider | Model Format |
|---|---|
| Anthropic | anthropic:claude-sonnet-4-20250514 |
| OpenAI | openai:gpt-4o |
| Azure OpenAI | azure:gpt-4o |
| AWS Bedrock | bedrock:anthropic.claude-3-sonnet |
google:gemini-1.5-pro |
Per-Request Model Selection
Switch models dynamically without restarting:
# Use Claude for complex reasoning
result = await agent.run("Analyze this data...", model="anthropic:claude-sonnet-4-20250514")
# Use GPT-4 for creative tasks
result = await agent.run("Write a poem...", model="openai:gpt-4o")
Future Framework Support
Community-Driven
We're open to expanding support for other agent frameworks based on community feedback:
- Google ADK — Google's Agent Development Kit
- LangChain — Popular Python agent framework
- CrewAI — Multi-agent orchestration
- AutoGen — Microsoft's agent framework
Share your feedback on which frameworks you'd like to see supported!