Human-in-the-loop

The Human-in-the-Loop (HITL) middleware lets you add human oversight to agent tool calls. When a model proposes an action that might require review — for example, writing to a file or executing SQL — the middleware can pause execution and wait for a decision.

It does this by checking each tool call against a configurable policy. If intervention is needed, the middleware issues an interrupt that halts execution. The graph state is saved using LangGraph’s persistence layer, so execution can pause safely and resume later.

A human decision then determines what happens next: the action can be approved as-is (approve), modified before running (edit), or rejected with feedback (reject).

Interrupt decision types

The middleware defines three built-in ways a human can respond to an interrupt:

Decision Type	Description	Example Use Case
✅ approve	The action is approved as-is and executed without changes.	Send an email draft exactly as written
✏️ edit	The tool call is executed with modifications.	Change the recipient before sending an email
❌ reject	The tool call is rejected, with an explanation added to the conversation.	Reject an email draft and explain how to rewrite it

The available decision types for each tool depend on the policy you configure in interrupt_on. When multiple tool calls are paused at the same time, each action requires a separate decision. Decisions must be provided in the same order as the actions appear in the interrupt request.

When editing tool arguments, make changes conservatively. Significant modifications to the original arguments may cause the model to re-evaluate its approach and potentially execute the tool multiple times or take unexpected actions.

Configuring interrupts

To use HITL, add the middleware to the agent’s middleware list when creating the agent.

You configure it with a mapping of tool actions to the decision types that are allowed for each action. The middleware will interrupt execution when a tool call matches an action in the mapping.

from langchain.agents import create_agent
from langchain.agents.middleware import HumanInTheLoopMiddleware 
from langgraph.checkpoint.memory import InMemorySaver 


agent = create_agent(
    model="gpt-4.1",
    tools=[write_file_tool, execute_sql_tool, read_data_tool],
    middleware=[
        HumanInTheLoopMiddleware( 
            interrupt_on={
                "write_file": True,  # All decisions (approve, edit, reject) allowed
                "execute_sql": {"allowed_decisions": ["approve", "reject"]},  # No editing allowed
                # Safe operation, no approval needed
                "read_data": False,
            },
            # Prefix for interrupt messages - combined with tool name and args to form the full message
            # e.g., "Tool execution pending approval: execute_sql with query='DELETE FROM...'"
            # Individual tools can override this by specifying a "description" in their interrupt config
            description_prefix="Tool execution pending approval",
        ),
    ],
    # Human-in-the-loop requires checkpointing to handle interrupts.
    # In production, use a persistent checkpointer like AsyncPostgresSaver.
    checkpointer=InMemorySaver(),  
)

You must configure a checkpointer to persist the graph state across interrupts. In production, use a persistent checkpointer like AsyncPostgresSaver. For testing or prototyping, use InMemorySaver.

When invoking the agent, pass a config that includes the thread ID to associate execution with a conversation thread. See the LangGraph interrupts documentation for details.

Configuration options

interrupt_on · dict · required

Mapping of tool names to approval configs. Values can be True (interrupt with default config), False (auto-approve), or an InterruptOnConfig object.

description_prefix · string · default: "Tool execution requires approval"

Prefix for action request descriptions

InterruptOnConfig options:

allowed_decisions · list[string]

List of allowed decisions: 'approve', 'edit', or 'reject'

description · string | callable

Static string or callable function for custom description

Responding to interrupts

When you invoke the agent, it runs until it either completes or an interrupt is raised. An interrupt is triggered when a tool call matches the policy you configured in interrupt_on. In that case, the invocation result will include an __interrupt__ field with the actions that require review. You can then present those actions to a reviewer and resume execution once decisions are provided.

from langgraph.types import Command

# Human-in-the-loop leverages LangGraph's persistence layer.
# You must provide a thread ID to associate the execution with a conversation thread,
# so the conversation can be paused and resumed (as is needed for human review).
config = {"configurable": {"thread_id": "some_id"}} 
# Run the graph until the interrupt is hit.
result = agent.invoke(
    {
        "messages": [
            {
                "role": "user",
                "content": "Delete old records from the database",
            }
        ]
    },
    config=config 
)

# The interrupt contains the full HITL request with action_requests and review_configs
print(result['__interrupt__'])
# > [
# >    Interrupt(
# >       value={
# >          'action_requests': [
# >             {
# >                'name': 'execute_sql',
# >                'arguments': {'query': 'DELETE FROM records WHERE created_at < NOW() - INTERVAL \'30 days\';'},
# >                'description': 'Tool execution pending approval\n\nTool: execute_sql\nArgs: {...}'
# >             }
# >          ],
# >          'review_configs': [
# >             {
# >                'action_name': 'execute_sql',
# >                'allowed_decisions': ['approve', 'reject']
# >             }
# >          ]
# >       }
# >    )
# > ]


# Resume with approval decision
agent.invoke(
    Command( 
        resume={"decisions": [{"type": "approve"}]}  # or "reject"
    ), 
    config=config # Same thread ID to resume the paused conversation
)

Decision types

✅ approve

Use approve to approve the tool call as-is and execute it without changes.

agent.invoke(
    Command(
        # Decisions are provided as a list, one per action under review.
        # The order of decisions must match the order of actions
        # listed in the `__interrupt__` request.
        resume={
            "decisions": [
                {
                    "type": "approve",
                }
            ]
        }
    ),
    config=config  # Same thread ID to resume the paused conversation
)

✏️ edit

Use edit to modify the tool call before execution. Provide the edited action with the new tool name and arguments.

agent.invoke(
    Command(
        # Decisions are provided as a list, one per action under review.
        # The order of decisions must match the order of actions
        # listed in the `__interrupt__` request.
        resume={
            "decisions": [
                {
                    "type": "edit",
                    # Edited action with tool name and args
                    "edited_action": {
                        # Tool name to call.
                        # Will usually be the same as the original action.
                        "name": "new_tool_name",
                        # Arguments to pass to the tool.
                        "args": {"key1": "new_value", "key2": "original_value"},
                    }
                }
            ]
        }
    ),
    config=config  # Same thread ID to resume the paused conversation
)

When editing tool arguments, make changes conservatively. Significant modifications to the original arguments may cause the model to re-evaluate its approach and potentially execute the tool multiple times or take unexpected actions.

❌ reject

Use reject to reject the tool call and provide feedback instead of execution.

agent.invoke(
    Command(
        # Decisions are provided as a list, one per action under review.
        # The order of decisions must match the order of actions
        # listed in the `__interrupt__` request.
        resume={
            "decisions": [
                {
                    "type": "reject",
                    # An explanation about why the action was rejected
                    "message": "No, this is wrong because ..., instead do this ...",
                }
            ]
        }
    ),
    config=config  # Same thread ID to resume the paused conversation
)

The message is added to the conversation as feedback to help the agent understand why the action was rejected and what it should do instead.

---

Multiple decisions

When multiple actions are under review, provide a decision for each action in the same order as they appear in the interrupt:

{
    "decisions": [
        {"type": "approve"},
        {
            "type": "edit",
            "edited_action": {
                "name": "tool_name",
                "args": {"param": "new_value"}
            }
        },
        {
            "type": "reject",
            "message": "This action is not allowed"
        }
    ]
}

Streaming with human-in-the-loop

You can use stream() instead of invoke() to get real-time updates while the agent runs and handles interrupts. Use stream_mode=['updates', 'messages'] to stream both agent progress and LLM tokens.

from langgraph.types import Command

config = {"configurable": {"thread_id": "some_id"}}

# Stream agent progress and LLM tokens until interrupt
for mode, chunk in agent.stream(
    {"messages": [{"role": "user", "content": "Delete old records from the database"}]},
    config=config,
    stream_mode=["updates", "messages"],  
):
    if mode == "messages":
        # LLM token
        token, metadata = chunk
        if token.content:
            print(token.content, end="", flush=True)
    elif mode == "updates":
        # Check for interrupt
        if "__interrupt__" in chunk:
            print(f"\n\nInterrupt: {chunk['__interrupt__']}")

# Resume with streaming after human decision
for mode, chunk in agent.stream(
    Command(resume={"decisions": [{"type": "approve"}]}),
    config=config,
    stream_mode=["updates", "messages"],
):
    if mode == "messages":
        token, metadata = chunk
        if token.content:
            print(token.content, end="", flush=True)

See the Streaming guide for more details on stream modes.

Execution lifecycle

The middleware defines an after_model hook that runs after the model generates a response but before any tool calls are executed:

1. The agent invokes the model to generate a response.
2. The middleware inspects the response for tool calls.
3. If any calls require human input, the middleware builds a HITLRequest with action_requests and review_configs and calls interrupt.
4. The agent waits for human decisions.
5. Based on the HITLResponse decisions, the middleware executes approved or edited calls, synthesizes ToolMessage’s for rejected calls, and resumes execution.

Custom HITL logic

For more specialized workflows, you can build custom HITL logic directly using the interrupt primitive and middleware abstraction.

Review the execution lifecycle above to understand how to integrate interrupts into the agent’s operation.

---

Edit this page on GitHub or file an issue.

Connect these docs to Claude, VSCode, and more via MCP for real-time answers.