# Sub-Agents

> Decompose work across multiple specialized agents with a visible delegation log.

<!-- interactive demo: subagents -->

## What is this? [#what-is-this]

Sub-agents are the canonical multi-agent pattern: a top-level
**supervisor** LLM orchestrates one or more specialized **sub-agents**
by exposing each of them as a tool. The supervisor decides what to
delegate, the sub-agents do their narrow job, and their results flow
back up to the supervisor's next step.

This is fundamentally the same shape as tool-calling, but each "tool"
is itself a full-blown agent with its own system prompt and (often) its
own tools, memory, and model.

## When should I use this? [#when-should-i-use-this]

Reach for sub-agents when a task has distinct specialized sub-tasks
that each benefit from their own focus:

- **Research → Write → Critique** pipelines, where each stage needs a
  different system prompt and temperature.
- **Router + specialists**, where one agent classifies the request and
  dispatches to the right expert.
- **Divide-and-conquer**, for any problem that fits cleanly into parallel
  or sequential sub-problems.

The example below uses the Research → Write → Critique shape as the
canonical example.

## Setting up sub-agents [#setting-up-sub-agents]

Define one agent per role, each with its own system prompt:

```python title="src/agents/subagents.py"
from __future__ import annotations

import logging
import uuid
from typing import Any, Literal

from pydantic import BaseModel, Field
from pydantic_ai import Agent, RunContext
from pydantic_ai.ui import StateDeps
from pydantic_ai.models.openai import OpenAIResponsesModel

logger = logging.getLogger(__name__)


SubAgentName = Literal["research_agent", "writing_agent", "critique_agent"]
DelegationStatus = Literal["running", "completed", "failed"]


# ── Shared state ────────────────────────────────────────────────────


class Delegation(BaseModel):
    """One sub-agent invocation, surfaced in the UI's delegation log."""

    id: str
    sub_agent: str  # SubAgentName at runtime; widened so model_dump round-trips
    task: str
    status: str  # DelegationStatus at runtime
    result: str = ""


class SubagentsState(BaseModel):
    """Shared state. ``delegations`` is rendered as a live log in the UI."""

    delegations: list[Delegation] = Field(default_factory=list)


# ── Sub-agents (real PydanticAI Agents) ─────────────────────────────


# Each sub-agent is a full-fledged ``Agent(model=..., system_prompt=...)``
# with its own system prompt. They don't share memory or tools with the
# supervisor — the supervisor only sees their return value.
_SUB_MODEL = OpenAIResponsesModel("gpt-6-luna")

_research_agent: Agent[None, str] = Agent(
    model=_SUB_MODEL,
    system_prompt=(
        "You are a research sub-agent. Given a topic, produce a concise "
        "bulleted list of 3-5 key facts. No preamble, no closing."
    ),
)

_writing_agent: Agent[None, str] = Agent(
    model=_SUB_MODEL,
    system_prompt=(
        "You are a writing sub-agent. Given a brief and optional source "
        "facts, produce a polished 1-paragraph draft. Be clear and "
        "concrete. No preamble."
    ),
)

_critique_agent: Agent[None, str] = Agent(
    model=_SUB_MODEL,
    system_prompt=(
        "You are an editorial critique sub-agent. Given a draft, give "
        "2-3 crisp, actionable critiques. No preamble."
    ),
)
```

Each sub-agent is an isolated agent call with its own model, system
prompt, and optional tools. They don't share memory or tools with the
supervisor; the supervisor only ever sees what the sub-agent returns.

Keep sub-agent system prompts narrow and focused. The point of this pattern
is that each one does one thing well. If a sub-agent needs to know
the whole user context to do its job, that's a signal the boundary is
wrong.

## Exposing sub-agents as tools [#exposing-sub-agents-as-tools]

The supervisor delegates by calling tools. Each delegation tool is a thin
wrapper around a specialized agent call that:

1. Runs the sub-agent on the supplied `task` string.
2. Records the delegation into a `delegations` slot in [shared agent
   state](/vue/pydantic-ai/shared-state) (so the UI can render a live log). In the Built-in Agent
   example, the tool code below does not write this entry. The agent
   loop writes it: after each delegation tool result, the loop emits a
   `STATE_DELTA` event that sets `/delegations`.
3. Returns the sub-agent's final message as the tool result, which the
   supervisor sees on its next turn.

```python title="src/agents/subagents.py"
from __future__ import annotations

import logging
import uuid
from typing import Any, Literal

from pydantic import BaseModel, Field
from pydantic_ai import Agent, RunContext
from pydantic_ai.ui import StateDeps
from pydantic_ai.models.openai import OpenAIResponsesModel

logger = logging.getLogger(__name__)


SubAgentName = Literal["research_agent", "writing_agent", "critique_agent"]
DelegationStatus = Literal["running", "completed", "failed"]


# ── Shared state ────────────────────────────────────────────────────


class Delegation(BaseModel):
    """One sub-agent invocation, surfaced in the UI's delegation log."""

    id: str
    sub_agent: str  # SubAgentName at runtime; widened so model_dump round-trips
    task: str
    status: str  # DelegationStatus at runtime
    result: str = ""


class SubagentsState(BaseModel):
    """Shared state. ``delegations`` is rendered as a live log in the UI."""

    delegations: list[Delegation] = Field(default_factory=list)


# ── Sub-agents (real PydanticAI Agents) ─────────────────────────────


# Each sub-agent is a full-fledged ``Agent(model=..., system_prompt=...)``
# with its own system prompt. They don't share memory or tools with the
# supervisor — the supervisor only sees their return value.
_SUB_MODEL = OpenAIResponsesModel("gpt-6-luna")

_research_agent: Agent[None, str] = Agent(
    model=_SUB_MODEL,
    system_prompt=(
        "You are a research sub-agent. Given a topic, produce a concise "
        "bulleted list of 3-5 key facts. No preamble, no closing."
    ),
)

_writing_agent: Agent[None, str] = Agent(
    model=_SUB_MODEL,
    system_prompt=(
        "You are a writing sub-agent. Given a brief and optional source "
        "facts, produce a polished 1-paragraph draft. Be clear and "
        "concrete. No preamble."
    ),
)

_critique_agent: Agent[None, str] = Agent(
    model=_SUB_MODEL,
    system_prompt=(
        "You are an editorial critique sub-agent. Given a draft, give "
        "2-3 crisp, actionable critiques. No preamble."
    ),
)


async def _invoke_sub_agent(sub_agent: Agent[None, str], task: str) -> str:
    """Run a sub-agent on ``task`` and return its final text output.

    Uses the async ``Agent.run`` API rather than ``run_sync`` because the
    supervisor itself executes inside a running event loop (AG-UI awaits
    ``agent.run``); ``run_sync`` from inside a running loop raises
    ``RuntimeError: This event loop is already running``.
    """
    result = await sub_agent.run(task)
    output: Any = result.output
    return str(output) if output is not None else ""


# ── Supervisor ──────────────────────────────────────────────────────


_SUPERVISOR_PROMPT = (
    "You are a supervisor agent that coordinates three specialized "
    "sub-agents to produce high-quality deliverables.\n\n"
    "Available sub-agents (call them as tools):\n"
    "  - research_agent: gathers facts on a topic.\n"
    "  - writing_agent: turns facts + a brief into a polished draft.\n"
    "  - critique_agent: reviews a draft and suggests improvements.\n\n"
    "For most non-trivial user requests, delegate in sequence: research -> "
    "write -> critique. Pass the relevant facts/draft through the `task` "
    "argument of each tool. Each tool returns the sub-agent's output as a "
    "string. Keep your own messages short — explain the plan once, "
    "delegate, then return a concise summary once done. The UI shows the "
    "user a live log of every sub-agent delegation, including the "
    "in-flight 'running' state."
)


agent = Agent(
    model=OpenAIResponsesModel("gpt-6-luna"),
    deps_type=StateDeps[SubagentsState],
    system_prompt=_SUPERVISOR_PROMPT,
)


def _append_running(
    ctx: RunContext[StateDeps[SubagentsState]],
    *,
    sub_agent: SubAgentName,
    task: str,
) -> str:
    """Append a ``running`` delegation entry and return its id."""
    entry = Delegation(
        id=str(uuid.uuid4()),
        sub_agent=sub_agent,
        task=task,
        status="running",
        result="",
    )
    ctx.deps.state.delegations = [*ctx.deps.state.delegations, entry]
    return entry.id


def _finalise(
    ctx: RunContext[StateDeps[SubagentsState]],
    *,
    entry_id: str,
    status: DelegationStatus,
    result: str,
) -> None:
    """Mutate the delegation entry with ``entry_id`` to its terminal state.

    If the entry has gone missing (e.g. another part of the system replaced
    ``state.delegations`` mid-turn) we log a warning and skip rather than
    appending a synthetic entry — same defensive choice as google-adk's
    ``_update_delegation``.
    """
    delegations = list(ctx.deps.state.delegations)
    for idx, entry in enumerate(delegations):
        if entry.id == entry_id:
            delegations[idx] = entry.model_copy(
                update={"status": status, "result": result}
            )
            ctx.deps.state.delegations = delegations
            return
    logger.warning(
        "subagents: delegation entry %s missing on update — final %s state "
        "(result_length=%d) will not be rendered",
        entry_id,
        status,
        len(result),
    )


# ── Delegation tools ────────────────────────────────────────────────


async def _delegate(
    ctx: RunContext[StateDeps[SubagentsState]],
    *,
    sub_agent: SubAgentName,
    sub_agent_obj: Agent[None, str],
    task: str,
) -> str:
    """Common delegation flow: append running → invoke → finalise.

    Returns the sub-agent's output text so the supervisor LLM can read it
    on its next step. State mutations (running entry + final entry) are
    written through ``ctx.deps.state``; PydanticAI's AG-UI bridge syncs
    those back to the frontend at end-of-turn so the delegation log
    re-renders automatically.
    """
    entry_id = _append_running(ctx, sub_agent=sub_agent, task=task)
    try:
        result = await _invoke_sub_agent(sub_agent_obj, task)
    except Exception as exc:  # noqa: BLE001 — surface failure to supervisor
        logger.exception("subagents: %s failed", sub_agent)
        message = (
            f"sub-agent {sub_agent} failed: {exc.__class__.__name__} "
            "(see server logs for details)"
        )
        _finalise(ctx, entry_id=entry_id, status="failed", result=message)
        return message
    _finalise(ctx, entry_id=entry_id, status="completed", result=result)
    return result


# Each ``@agent.tool`` wraps a sub-agent invocation. The supervisor LLM
# "calls" these tools to delegate work; each call asynchronously runs the
# matching sub-agent, records the delegation into shared state, and
# returns the sub-agent's output as a string the supervisor can read on
# its next step.
@agent.tool
async def research_agent(
    ctx: RunContext[StateDeps[SubagentsState]],
    task: str,
) -> str:
    """Delegate a research task to the research sub-agent.

    Use for: gathering facts, background, definitions, statistics.
    Returns a bulleted list of key facts.
    """
    return await _delegate(
        ctx,
        sub_agent="research_agent",
        sub_agent_obj=_research_agent,
        task=task,
    )


@agent.tool
async def writing_agent(
    ctx: RunContext[StateDeps[SubagentsState]],
    task: str,
) -> str:
    """Delegate a drafting task to the writing sub-agent.

    Use for: producing a polished paragraph, draft, or summary. Pass
    relevant facts from prior research inside ``task``.
    """
    return await _delegate(
        ctx,
        sub_agent="writing_agent",
        sub_agent_obj=_writing_agent,
        task=task,
    )


@agent.tool
async def critique_agent(
    ctx: RunContext[StateDeps[SubagentsState]],
    task: str,
) -> str:
    """Delegate a critique task to the critique sub-agent.

    Use for: reviewing a draft and suggesting concrete improvements.
    """
    return await _delegate(
        ctx,
        sub_agent="critique_agent",
        sub_agent_obj=_critique_agent,
        task=task,
    )

```

If the excerpt uses a header-forwarding helper, such as `forwardingFetch` or `HeaderForwardingMiddleware`, leave it out of your app. The showcase uses it only to send request headers to its test server.

This is where CopilotKit's shared-state channel earns its keep: the
supervisor's tool calls mutate `delegations` as they happen, and the
frontend renders every new entry live.

<Callout type="warn">
  Give every delegation a stable `id` and merge new entries by that `id`. The
  client sends its copy of shared state back as run input on every run. A slot
  that appends everything it receives, such as a LangGraph
  `Annotated[list, operator.add]` reducer, adds those entries again. Then the
  log doubles when a thread continues. The LangGraph excerpt above uses
  `operator.add`, so in your app, write a reducer that merges by `id`.
</Callout>

## Rendering a live delegation log [#rendering-a-live-delegation-log]

On the frontend, the delegation log is a reactive render of the
`delegations` slot.

To stream each sub-agent's output token by token inside its log entry, see
[State streaming](/vue/pydantic-ai/shared-state/streaming).

The log grows in real time as the supervisor delegates. The user sees each
step while it happens, not a long spinner.
