State Streaming

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

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

Stream partial state updates while Claude responds

For streaming state, parse the agent's structured deltas as they arrive and emit CopilotKit state updates before the final message is complete. This branch runs inside the streamed tool-argument handler.

agent.py
                            if current_tool_name == "write_document":
                                streamed_document = _partial_json_string_property(
                                    current_tool_args,
                                    "document",
                                )
                                if (
                                    streamed_document is not None
                                    and streamed_document != last_streamed_document
                                ):
                                    state.document = streamed_document
                                    last_streamed_document = streamed_document
                                    yield encoder.encode(
                                        StateSnapshotEvent(
                                            type=EventType.STATE_SNAPSHOT,
                                            snapshot=state.model_dump(),
                                        )
                                    )

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.

state_streaming_backend.py
import json
from collections.abc import AsyncIterator
from typing import Any

from ag_ui.core import EventType, StateSnapshotEvent
from ag_ui.encoder import EventEncoder
from pydantic import BaseModel


class AgentState(BaseModel):
    document: str = ""


WRITE_DOCUMENT_TOOL_SCHEMA: dict[str, Any] = {
    "name": "write_document",
    "description": "Write a document into shared agent state.",
    "input_schema": {
        "type": "object",
        "properties": {
            "document": {
                "type": "string",
                "description": "The full document text to render in shared state.",
            },
        },
        "required": ["document"],
    },
}


def _partial_json_string_property(source: str, key: str) -> str | None:
    """Return the current value of a streamed JSON string property."""
    marker = json.dumps(key)
    key_pos = source.find(marker)
    if key_pos < 0:
        return None
    colon_pos = source.find(":", key_pos)
    if colon_pos < 0:
        # The colon hasn't streamed yet (e.g. source is `{"document`); bail out
        # rather than letting find(":") return -1 and matching the key's own
        # opening quote, which would report the key name as the value.
        return None
    value_start = source.find('"', colon_pos + 1)
    if value_start < 0:
        return None

    raw_chars: list[str] = []
    escaped = False
    for char in source[value_start + 1 :]:
        if escaped:
            raw_chars.append("\\" + char)
            escaped = False
        elif char == "\\":
            escaped = True
        elif char == '"':
            break
        else:
            raw_chars.append(char)

    try:
        return json.loads(f'"{"".join(raw_chars)}"')
    except json.JSONDecodeError:
        return None


async def stream_document_state(
    anthropic_stream: AsyncIterator[Any],
    state: AgentState,
) -> AsyncIterator[str]:
    encoder = EventEncoder()
    current_tool_name: str | None = None
    current_tool_args = ""
    last_streamed_document = state.document

    async for event in anthropic_stream:
        if (
            event.type == "content_block_start"
            and event.content_block.type == "tool_use"
        ):
            current_tool_name = event.content_block.name
            current_tool_args = ""
            continue

        if (
            event.type != "content_block_delta"
            or event.delta.type != "input_json_delta"
        ):
            continue

        current_tool_args += event.delta.partial_json
        if current_tool_name != "write_document":
            continue

        streamed_document = _partial_json_string_property(
            current_tool_args,
            "document",
        )
        if streamed_document is None or streamed_document == last_streamed_document:
            continue

        state.document = streamed_document
        last_streamed_document = streamed_document
        yield encoder.encode(
            StateSnapshotEvent(
                type=EventType.STATE_SNAPSHOT,
                snapshot=state.model_dump(),
            )
        )

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?