State Streaming

Stream partial agent state updates to the UI while a tool call is still running.

Shared State: Streaming

Waking up the demo

What is this?

By default, agent state only updates between backend checkpoints, so a long-running tool call (writing a full document, drafting an email) appears to the UI as one big burst at the end. For agent-native apps, that feels broken: users expect to watch the output materialise.

State streaming copies one tool argument into an agent state key while the model writes the argument. The UI, subscribed via useAgent, re-renders every token.

When should I use this?

Use state streaming whenever a tool's output is long-form text or a growing structured value and you want the user to see it assemble in real time. Common shapes:

  • A collaborative writing agent that emits a document
  • A research agent that accumulates a list of findings
  • A planning agent that builds up a step-by-step plan

The backend: one streaming state mapping

Install the LangGraph Python SDK

uv add copilotkit

Add state streaming middleware

StateStreamingMiddleware maps a tool argument to an agent state key and forwards partial tool-call arguments as state updates while the model is still generating.

shared_state_streaming.py
import uuid

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

from copilotkit import (
    CopilotKitMiddleware,
    StateItem,
    StateStreamingMiddleware,
)


class AgentState(BaseAgentState):
    """Shared state. `document` is streamed token-by-token."""

    document: str


@tool
def write_document(document: str, runtime: ToolRuntime) -> Command:
    """Write a document for the user.

    Always call this tool when the user asks you to write or draft
    something of any length (an essay, poem, email, summary, etc.).
    The `document` argument is streamed *per token* into shared agent
    state under the `document` key, so the UI can render it as it is
    generated.
    """
    return Command(
        update={
            "document": document,
            "messages": [
                ToolMessage(
                    content="Document written to shared state.",
                    name="write_document",
                    id=str(uuid.uuid4()),
                    tool_call_id=runtime.tool_call_id,
                )
            ],
        }
    )


graph = create_agent(
    model=ChatOpenAI(reasoning_effort="none", model="gpt-6-luna"),
    tools=[write_document],
    middleware=[
        CopilotKitMiddleware(),
        # Forward every token of write_document's `document` argument
        # straight into state["document"] while the tool call is still
        # streaming. Without this, `document` would only update once
        # the tool call completes.
        #
        # NOTE: the frontend `usePredictStateSubscription` hook indexes
        # the (partial-JSON-parsed) tool args by `state_key`, so the
        # tool's argument name MUST match `state_key` ("document") for
        # per-token deltas to land in `state.document`.
        StateStreamingMiddleware(
            StateItem(
                state_key="document",
                tool="write_document",
                tool_argument="document",
            )
        ),
    ],
    state_schema=AgentState,
    system_prompt=(
        "You are a collaborative writing assistant. Whenever the user asks "
        "you to write, draft, or revise any piece of text, ALWAYS call the "
        "`write_document` tool with the full content as a single string in "
        "the `document` argument. Never paste the document into a chat "
        "message directly — the document belongs in shared state and the "
        "UI renders it live as you type."
    ),
)

Using a custom graph instead of create_agent

StateStreamingMiddleware is a create_agent middleware. In a graph where you define the nodes and edges yourself, map the tool argument with copilotkit_customize_config and pass the returned config to the model call:

agent.py
from copilotkit import CopilotKitState
from copilotkit.langgraph import copilotkit_customize_config
from langchain_core.messages import SystemMessage
from langchain_core.runnables import RunnableConfig

class AgentState(CopilotKitState):
    observed_steps: list[str]

async def chat_node(state: AgentState, config: RunnableConfig):
    streaming_config = copilotkit_customize_config(
        config,
        emit_intermediate_state=[
            {
                "state_key": "observed_steps",
                "tool": "step_progress_tool",
                "tool_argument": "steps",
            },
        ]
    )

    model_with_tools = model.bind_tools(
        [
            *state["copilotkit"]["actions"],
            *tools,
        ],
        parallel_tool_calls=False,
    )

    response = await model_with_tools.ainvoke(
        [
            SystemMessage(
                content="You are a task performer. Report your steps "
                "using step_progress_tool."
            ),
            *state["messages"],
        ],
        streaming_config,
    )
    return {"messages": [response]}

To push progress from your own code instead of a tool call, call copilotkit_emit_state with the state you want the UI to show:

agent.py
from copilotkit.langgraph import copilotkit_emit_state

for step in steps:
    state["observed_steps"] = state.get("observed_steps", []) + [step]
    await copilotkit_emit_state(config, state)

Streamed values are previews. When the node finishes, the state it returns replaces them, so return every key you want to keep, or the UI loses the streamed value.

The backend pattern is always the same: map one streaming tool argument to one shared-state key. Each framework does this in its own way:

  • LangGraph Python: StateStreamingMiddleware with StateItem(...) entries.
  • LangGraph TypeScript: copilotkitCustomizeConfig with an emitIntermediateState mapping.
  • Direct SDK adapters: parse partial tool arguments in the streaming loop, and emit STATE_SNAPSHOT whenever the mapped value changes.
  • LlamaIndex and Pydantic AI: there is no per-token mapping. A backend tool sends a state snapshot after each step or tool call instead (see the setup above).

When the LLM streams that argument, CopilotKit writes every partial value into shared state before the tool even finishes executing.

shared_state_streaming.py
import uuid

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

from copilotkit import (
    CopilotKitMiddleware,
    StateItem,
    StateStreamingMiddleware,
)


class AgentState(BaseAgentState):
    """Shared state. `document` is streamed token-by-token."""

    document: str


@tool
def write_document(document: str, runtime: ToolRuntime) -> Command:
    """Write a document for the user.

    Always call this tool when the user asks you to write or draft
    something of any length (an essay, poem, email, summary, etc.).
    The `document` argument is streamed *per token* into shared agent
    state under the `document` key, so the UI can render it as it is
    generated.
    """
    return Command(
        update={
            "document": document,
            "messages": [
                ToolMessage(
                    content="Document written to shared state.",
                    name="write_document",
                    id=str(uuid.uuid4()),
                    tool_call_id=runtime.tool_call_id,
                )
            ],
        }
    )


graph = create_agent(
    model=ChatOpenAI(reasoning_effort="none", model="gpt-6-luna"),
    tools=[write_document],
    middleware=[
        CopilotKitMiddleware(),
        # Forward every token of write_document's `document` argument
        # straight into state["document"] while the tool call is still
        # streaming. Without this, `document` would only update once
        # the tool call completes.
        #
        # NOTE: the frontend `usePredictStateSubscription` hook indexes
        # the (partial-JSON-parsed) tool args by `state_key`, so the
        # tool's argument name MUST match `state_key` ("document") for
        # per-token deltas to land in `state.document`.
        StateStreamingMiddleware(
            StateItem(
                state_key="document",
                tool="write_document",
                tool_argument="document",
            )
        ),
    ],
    state_schema=AgentState,
    system_prompt=(
        "You are a collaborative writing assistant. Whenever the user asks "
        "you to write, draft, or revise any piece of text, ALWAYS call the "
        "`write_document` tool with the full content as a single string in "
        "the `document` argument. Never paste the document into a chat "
        "message directly — the document belongs in shared state and the "
        "UI renders it live as you type."
    ),
)

A few things to note:

  • The state key must exist in your agent state (document in this demo).
  • The tool and argument names must match the exact LLM-facing tool call you want to forward (write_document.document here).
  • Streamed values are previews. When the step ends, the state the agent saves replaces them, so save the final value when the tool finishes. On CrewAI, LlamaIndex and Pydantic AI, write the final value to the agent state yourself, or it is overwritten when the step returns.

The frontend: useAgent + OnStateChanged

The UI side is the same as any other shared-state subscription: useAgent with OnStateChanged gives you a reactive agent.state. Add OnRunStatusChanged if you want a "LIVE" / "done" indicator.

page.tsx
  // Subscribe to BOTH state changes and run-status changes. The former
  // drives the per-token document rerender; the latter toggles the
  // "LIVE" badge when the agent starts / stops.
  const { agent } = useAgent({
    agentId: "shared-state-streaming",
    updates: [UseAgentUpdate.OnStateChanged, UseAgentUpdate.OnRunStatusChanged],
  });

From there, agent.state.document is just a string that grows on every token, and agent.isRunning tells you whether to show a streaming indicator.

Try it

Ask the agent to write a short poem. The document text grows in your UI while the agent writes. When the run ends, agent.isRunning is false, so your streaming indicator turns off.

How is this guide?