LangGraph middleware¶

Middleware adapts model requests before provider invocation. AI Hub currently uses it for:

• tenant/user-aware dynamic prompts;
• active-agent and tool filtering;
• message-count and token-count history limits;
• persisted conversation summarization;
• cleanup of tool-call history and specialist-specific context.

Agents are created with langchain.agents.create_agent and receive middleware instances or decorated dynamic-prompt functions:

from langchain.agents import create_agent

agent = create_agent(
    model=llm,
    tools=tools,
    middleware=[
        LimitChatHistoryMiddleware(max_messages=10),
        create_dynamic_prompt_for_task_helper_middleware,
    ],
)

• Dynamic prompts
• Chat history limits