Agent Context

Overview¶

AgentContext provides an easy way to access the agent as well as the pool in tool calls. It can get requested separately from Pydantic-AIs RunContext, which focuses on the Context of the actual agent run.

Basic Usage¶

Tools can request context injection by adding an AgentContext parameter:

async def my_tool(ctx: AgentContext[TDeps], arg: str) -> str:
    """Tool with access to context and dependencies."""
    # Access pool functionality
    sub_agent = ctx.pool.get_agent("helper")
    result = await sub_agent.run(arg)
    return do_something_with_result(result)

It is also possible to request both contexts, in this case the Pydantic-AI RunContext needs to appear first in the signature.

async def my_tool(run_ctx: RunContext, agent_ctx: AgentContext, arg: str) -> str:
    """Tool which requires both contexts."""