@agent | Respan Docs

Overview

The @agent decorator creates a span representing an autonomous agent. Agent spans set the workflow name context, so all nested tasks and tools appear grouped under this agent in the trace tree.

from respan import agent

Parameters

Parameter	Type	Default	Description
`name`	`str \| None`	Function name	Display name for the agent span.
`version`	`int \| None`	`None`	Version number for the agent.
`method_name`	`str \| None`	`None`	Required when decorating a class. Specifies which method to use as the entry point.
`processors`	`str \| List[str] \| None`	`None`	Route this span to specific named processors. See add_processor.
`export_filter`	`FilterParamDict \| None`	`None`	Filter dict to control which spans are exported. Uses AND logic — all conditions must match.

Function usage

from respan import Respan, agent, task, tool
from openai import OpenAI
respan = Respan(api_key="your-api-key")
client = OpenAI()
@tool(name="search")
def search(query: str):
    return f"Results for: {query}"
@task(name="think")
def think(context: str):
    response = client.chat.completions.create(
        model="gpt-4o-mini",
        messages=[{"role": "user", "content": context}],
    )
    return response.choices[0].message.content
@agent(name="research_agent")
def research_agent(topic: str):
    results = search(topic)
    return think(results)
print(research_agent("OpenTelemetry tracing"))

Class usage

from respan import Respan, agent, task
respan = Respan(api_key="your-api-key")
@agent(name="assistant", method_name="run")
class Assistant:
    @task(name="respond")
    def respond(self, input_text: str):
        return f"Response: {input_text}"
    def run(self):
        return self.respond("Hello")
print(Assistant().run())

Notes

Agent spans set the workflow name context — nested spans inherit the agent name for grouping
Use @agent for top-level autonomous entities and @task for individual operations within them