Use this pattern when each request is independent and you do not need cross-run memory.

• chat-style helper endpoints
• repository summarization
• one-off analysis or classification
• draft generation with no follow-up state

from loom import Agent, Model, Runtime

agent = Agent(
    model=Model.anthropic("claude-sonnet-4"),
    instructions="You are a concise product assistant.",
    runtime=Runtime.sdk(),
)

result = await agent.run("Summarize the release notes")
print(result.output)

• the API stays minimal
• there is no session lifecycle to manage
• the request boundary is explicit

• add Generation when output style matters
• move to Session when later runs must remember earlier ones
• add KnowledgeQuery when answers must be grounded in explicit evidence
• add Capability when the agent needs files, web, shell, MCP, or skills

• examples/01_hello_agent.py
• examples/00_agent_overview.py