# 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]

<Steps>
  <Step>
    ### Install the LangGraph Python SDK

    <Tabs className="p-0 m-0" groupId="python-pm" items={['uv', 'poetry', 'pip', 'conda']} default="uv">
    <Tab value="uv">
    ```bash
    uv add copilotkit
    ```
    </Tab>
    <Tab value="poetry">
    ```bash
    poetry add copilotkit
    ```
    </Tab>

    <Tab value="pip">

    ```bash
    pip install copilotkit --extra-index-url https://copilotkit.gateway.scarf.sh/simple/
    ```
    </Tab>

    <Tab value="conda">
    ```bash
    conda install copilotkit -c copilotkit-channel
    ```
    </Tab>

</Tabs>

  </Step>
  <Step>
    ### Wire CopilotKit middleware into your graph

    Sub-agents are wired as tools on a supervisor `create_agent` call. The
    supervisor's middleware list includes `CopilotKitMiddleware()` so the
    delegation log slot streams back to the UI through shared state.

~~~~python title="subagents.py"
import operator
import uuid
from typing import Annotated, Literal, TypedDict

from langchain.agents import AgentState as BaseAgentState, create_agent
from langchain.tools import ToolRuntime, tool
from langchain_core.messages import AIMessage, HumanMessage, ToolMessage
from langchain_openai import ChatOpenAI
from langgraph.types import Command

from copilotkit import CopilotKitMiddleware

from src.agents._header_forwarding_middleware import HeaderForwardingMiddleware


# ---------------------------------------------------------------------------
# Shared state
# ---------------------------------------------------------------------------


class Delegation(TypedDict):
    id: str
    sub_agent: Literal["research_agent", "writing_agent", "critique_agent"]
    task: str
    status: Literal["completed"]
    result: str


# Cap the supervisor → critique sub-agent loop at a single iteration.
# Without this, the supervisor LLM occasionally re-calls `critique_agent`
# repeatedly on the same draft (visible as stacking 🧐 cards in the
# chat). The critic only adds value once per draft, so we hard-stop
# after `_MAX_CRITIQUE_ITERATIONS` invocations and return a no-op
# result.
_MAX_CRITIQUE_ITERATIONS = 1


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

    `delegations` uses an `operator.add` reducer so that concurrent
    sub-agent emissions in the same supervisor step accumulate into a
    single list instead of conflicting (LangGraph would otherwise raise
    `INVALID_CONCURRENT_GRAPH_UPDATE` — "Can receive only one value per
    step. Use an Annotated key to handle multiple values.").
    """

    delegations: Annotated[list[Delegation], operator.add]


# ---------------------------------------------------------------------------
# Sub-agents (real LLM agents under the hood)
# ---------------------------------------------------------------------------

# Each sub-agent is a full-fledged `create_agent(...)` with its own
# system prompt. They don't share memory or tools with the supervisor —
# the supervisor only sees their return value.
_sub_model = ChatOpenAI(reasoning_effort="none", model="gpt-6-luna")

# Each sub-agent gets the minimal HeaderForwardingMiddleware so the
# inbound x-aimock-context (and other x-*) headers from the supervisor's
# inbound HTTP request propagate to the sub-agent's outbound LLM call.
# We intentionally do NOT attach the full CopilotKitMiddleware here —
# the supervisor handles App-Context / frontend-tool injection for the
# whole run, and adding it on sub-agents would double-inject prompt
# state. Header forwarding alone is enough to keep aimock matching.
_research_agent = create_agent(
    model=_sub_model,
    tools=[],
    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."
    ),
    middleware=[HeaderForwardingMiddleware()],
)

_writing_agent = create_agent(
    model=_sub_model,
    tools=[],
    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."
    ),
    middleware=[HeaderForwardingMiddleware()],
)

_critique_agent = create_agent(
    model=_sub_model,
    tools=[],
    system_prompt=(
        "You are an editorial critique sub-agent. Given a draft, give "
        "2-3 crisp, actionable critiques. No preamble."
    ),
    middleware=[HeaderForwardingMiddleware()],
)
~~~~

    See `src/agents/subagents.py` for the canonical supervisor + three
    sub-agents pattern with a live delegation log.

  </Step>
</Steps>

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

```python title="src/agents/subagents.py"
import operator
import uuid
from typing import Annotated, Literal, TypedDict

from langchain.agents import AgentState as BaseAgentState, create_agent
from langchain.tools import ToolRuntime, tool
from langchain_core.messages import AIMessage, HumanMessage, ToolMessage
from langchain_openai import ChatOpenAI
from langgraph.types import Command

from copilotkit import CopilotKitMiddleware

from src.agents._header_forwarding_middleware import HeaderForwardingMiddleware


# ---------------------------------------------------------------------------
# Shared state
# ---------------------------------------------------------------------------


class Delegation(TypedDict):
    id: str
    sub_agent: Literal["research_agent", "writing_agent", "critique_agent"]
    task: str
    status: Literal["completed"]
    result: str


# Cap the supervisor → critique sub-agent loop at a single iteration.
# Without this, the supervisor LLM occasionally re-calls `critique_agent`
# repeatedly on the same draft (visible as stacking 🧐 cards in the
# chat). The critic only adds value once per draft, so we hard-stop
# after `_MAX_CRITIQUE_ITERATIONS` invocations and return a no-op
# result.
_MAX_CRITIQUE_ITERATIONS = 1


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

    `delegations` uses an `operator.add` reducer so that concurrent
    sub-agent emissions in the same supervisor step accumulate into a
    single list instead of conflicting (LangGraph would otherwise raise
    `INVALID_CONCURRENT_GRAPH_UPDATE` — "Can receive only one value per
    step. Use an Annotated key to handle multiple values.").
    """

    delegations: Annotated[list[Delegation], operator.add]


# ---------------------------------------------------------------------------
# Sub-agents (real LLM agents under the hood)
# ---------------------------------------------------------------------------

# Each sub-agent is a full-fledged `create_agent(...)` with its own
# system prompt. They don't share memory or tools with the supervisor —
# the supervisor only sees their return value.
_sub_model = ChatOpenAI(reasoning_effort="none", model="gpt-6-luna")

# Each sub-agent gets the minimal HeaderForwardingMiddleware so the
# inbound x-aimock-context (and other x-*) headers from the supervisor's
# inbound HTTP request propagate to the sub-agent's outbound LLM call.
# We intentionally do NOT attach the full CopilotKitMiddleware here —
# the supervisor handles App-Context / frontend-tool injection for the
# whole run, and adding it on sub-agents would double-inject prompt
# state. Header forwarding alone is enough to keep aimock matching.
_research_agent = create_agent(
    model=_sub_model,
    tools=[],
    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."
    ),
    middleware=[HeaderForwardingMiddleware()],
)

_writing_agent = create_agent(
    model=_sub_model,
    tools=[],
    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."
    ),
    middleware=[HeaderForwardingMiddleware()],
)

_critique_agent = create_agent(
    model=_sub_model,
    tools=[],
    system_prompt=(
        "You are an editorial critique sub-agent. Given a draft, give "
        "2-3 crisp, actionable critiques. No preamble."
    ),
    middleware=[HeaderForwardingMiddleware()],
)
```

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/langgraph-python/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"
import operator
import uuid
from typing import Annotated, Literal, TypedDict

from langchain.agents import AgentState as BaseAgentState, create_agent
from langchain.tools import ToolRuntime, tool
from langchain_core.messages import AIMessage, HumanMessage, ToolMessage
from langchain_openai import ChatOpenAI
from langgraph.types import Command

from copilotkit import CopilotKitMiddleware

from src.agents._header_forwarding_middleware import HeaderForwardingMiddleware


# ---------------------------------------------------------------------------
# Shared state
# ---------------------------------------------------------------------------


class Delegation(TypedDict):
    id: str
    sub_agent: Literal["research_agent", "writing_agent", "critique_agent"]
    task: str
    status: Literal["completed"]
    result: str


# Cap the supervisor → critique sub-agent loop at a single iteration.
# Without this, the supervisor LLM occasionally re-calls `critique_agent`
# repeatedly on the same draft (visible as stacking 🧐 cards in the
# chat). The critic only adds value once per draft, so we hard-stop
# after `_MAX_CRITIQUE_ITERATIONS` invocations and return a no-op
# result.
_MAX_CRITIQUE_ITERATIONS = 1


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

    `delegations` uses an `operator.add` reducer so that concurrent
    sub-agent emissions in the same supervisor step accumulate into a
    single list instead of conflicting (LangGraph would otherwise raise
    `INVALID_CONCURRENT_GRAPH_UPDATE` — "Can receive only one value per
    step. Use an Annotated key to handle multiple values.").
    """

    delegations: Annotated[list[Delegation], operator.add]


# ---------------------------------------------------------------------------
# Sub-agents (real LLM agents under the hood)
# ---------------------------------------------------------------------------

# Each sub-agent is a full-fledged `create_agent(...)` with its own
# system prompt. They don't share memory or tools with the supervisor —
# the supervisor only sees their return value.
_sub_model = ChatOpenAI(reasoning_effort="none", model="gpt-6-luna")

# Each sub-agent gets the minimal HeaderForwardingMiddleware so the
# inbound x-aimock-context (and other x-*) headers from the supervisor's
# inbound HTTP request propagate to the sub-agent's outbound LLM call.
# We intentionally do NOT attach the full CopilotKitMiddleware here —
# the supervisor handles App-Context / frontend-tool injection for the
# whole run, and adding it on sub-agents would double-inject prompt
# state. Header forwarding alone is enough to keep aimock matching.
_research_agent = create_agent(
    model=_sub_model,
    tools=[],
    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."
    ),
    middleware=[HeaderForwardingMiddleware()],
)

_writing_agent = create_agent(
    model=_sub_model,
    tools=[],
    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."
    ),
    middleware=[HeaderForwardingMiddleware()],
)

_critique_agent = create_agent(
    model=_sub_model,
    tools=[],
    system_prompt=(
        "You are an editorial critique sub-agent. Given a draft, give "
        "2-3 crisp, actionable critiques. No preamble."
    ),
    middleware=[HeaderForwardingMiddleware()],
)


# Sentinel surfaced when a sub-agent run produces no usable text. Kept
# as a module-level constant so the harness probe (and any UI fallback)
# can match the exact phrase. The leading/trailing angle brackets keep
# it out of plausible LLM phrasing.
SUB_AGENT_EMPTY_SENTINEL = "<sub-agent produced no output>"


def _invoke_sub_agent(agent, task: str) -> str:
    """Run a sub-agent on `task` and return its final prose message."""
    result = agent.invoke({"messages": [HumanMessage(content=task)]})
    messages = result.get("messages", [])
    # Walk newest -> oldest so we pick the answer for THIS task, not a stale
    # intro. Skip empty AIMessages that only carry tool_calls.
    for msg in reversed(messages):
        if isinstance(msg, AIMessage):
            content = msg.content
            if isinstance(content, str) and content.strip():
                return content
            # Some providers stream content as a list of content blocks
            # (e.g. {"type": "text", "text": "..."}); concatenate the text.
            # The `isinstance(block.get("text"), str)` guard rejects
            # `{"type": "text", "text": null}` payloads — a known provider
            # quirk — that would otherwise crash `"".join(...)` with
            # `TypeError: sequence item N: expected str instance, NoneType found`.
            if isinstance(content, list):
                parts = [
                    block["text"]
                    for block in content
                    if isinstance(block, dict)
                    and block.get("type") == "text"
                    and isinstance(block.get("text"), str)
                ]
                joined = "".join(parts).strip()
                if joined:
                    return joined
    return SUB_AGENT_EMPTY_SENTINEL


def _delegation_update(
    sub_agent: str,
    task: str,
    result: str,
    tool_call_id: str,
) -> Command:
    """Append a completed delegation entry to shared state.

    Returns just the new entry (a one-element list). The reducer on
    `AgentState.delegations` is `operator.add`, which concatenates the
    new list with the prior state — so we must NOT echo back the
    existing delegations here, or they would be duplicated each step.
    """
    entry: Delegation = {
        "id": str(uuid.uuid4()),
        "sub_agent": sub_agent,  # type: ignore[typeddict-item]
        "task": task,
        "status": "completed",
        "result": result,
    }
    return Command(
        update={
            "delegations": [entry],
            "messages": [
                ToolMessage(
                    content=result,
                    name=sub_agent,
                    id=str(uuid.uuid4()),
                    tool_call_id=tool_call_id,
                )
            ],
        }
    )


# ---------------------------------------------------------------------------
# Supervisor tools (each tool delegates to one sub-agent)
# ---------------------------------------------------------------------------


# Each @tool wraps a sub-agent invocation. The supervisor LLM "calls"
# these tools to delegate work; each call synchronously runs the
# matching sub-agent, records the delegation into shared state, and
# returns the sub-agent's output as a ToolMessage the supervisor can
# read on its next step.
@tool
def research_agent(task: str, runtime: ToolRuntime) -> Command:
    """Delegate a research task to the research sub-agent.

    Use for: gathering facts, background, definitions, statistics.
    Returns a bulleted list of key facts.
    """
    result = _invoke_sub_agent(_research_agent, task)
    return _delegation_update("research_agent", task, result, runtime.tool_call_id)


@tool
def writing_agent(task: str, runtime: ToolRuntime) -> Command:
    """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`.
    """
    result = _invoke_sub_agent(_writing_agent, task)
    return _delegation_update("writing_agent", task, result, runtime.tool_call_id)


@tool
def critique_agent(task: str, runtime: ToolRuntime) -> Command:
    """Delegate a critique task to the critique sub-agent.

    Use for: reviewing a draft and suggesting concrete improvements.

    Capped at `_MAX_CRITIQUE_ITERATIONS` invocations per supervisor run
    — the supervisor LLM occasionally re-calls the critic in a loop and
    each rerun produces near-identical output, so additional calls are
    short-circuited with a no-op result that nudges the supervisor to
    finish.
    """
    state: AgentState = runtime.state  # type: ignore[assignment]
    delegations = state.get("delegations") or []
    prior_critiques = sum(
        1 for d in delegations if d.get("sub_agent") == "critique_agent"
    )
    if prior_critiques >= _MAX_CRITIQUE_ITERATIONS:
        # Short-circuit without appending another delegation entry — the
        # UI renders one card per delegation and we want exactly one
        # critic card per supervisor run, even if the LLM ignores the
        # system prompt and re-issues the call.
        skip_message = (
            "Critique already produced for this run. "
            "Stop calling critique_agent and return your final answer "
            "to the user now."
        )
        return Command(
            update={
                "messages": [
                    ToolMessage(
                        content=skip_message,
                        name="critique_agent",
                        id=str(uuid.uuid4()),
                        tool_call_id=runtime.tool_call_id,
                    )
                ],
            }
        )
    result = _invoke_sub_agent(_critique_agent, task)
    return _delegation_update("critique_agent", task, result, runtime.tool_call_id)

```

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/langgraph-python/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.
