langgraph-supervisor

What it is and what it does

langgraph-supervisor provides a Python library for building hierarchical multi-agent systems on top of LangGraph. It implements a supervisor pattern where a central agent orchestrates communication and task delegation to specialized worker agents. The supervisor makes routing decisions based on context and task requirements, controlling all message flow between agents.

The library handles the mechanics of agent coordination—tool-based handoffs, message history management, and multi-level hierarchies—so you can focus on defining specialized agents and their supervisor's decision logic. It integrates with LangGraph's streaming, memory, and human-in-the-loop capabilities. However, the library's own documentation now recommends using the supervisor pattern directly via tool-calling for most use cases, suggesting this library is best suited for users upgrading existing code or cases where its abstraction provides clear value.

Use it for:

• Route customer support queries to specialized agents (billing, technical, account management) based on intent
• Coordinate research and math agents to answer complex questions requiring both data lookup and calculation
• Build multi-level hierarchies where team supervisors report to a top-level supervisor for large organizations
• Manage conversation history and context across multiple specialized agents in a single workflow
• Add persistent memory and human approval steps to multi-agent workflows via LangGraph checkpointers

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

Builds hierarchical multi-agent systems where a supervisor agent coordinates and delegates tasks to specialized worker agents using LangGraph.

Yes, if you are building a hierarchical multi-agent system and prefer a structured abstraction over manual tool-calling. The library is actively maintained, has low install friction, and integrates well with LangGraph's ecosystem. However, review the library's own recommendation to use tool-calling directly for most new projects—install this primarily to upgrade existing code or when the supervisor abstraction clearly simplifies your architecture.

Install

langgraph-supervisor on PyPI

pip install langgraph-supervisor

uv add langgraph-supervisor

poetry add langgraph-supervisor

Installing langgraph-supervisor

Before you install

Low install friction with only two runtime dependencies (langgraph and langchain-core). Active maintenance with recent releases; last commit 2026-07-15. Requires Python >= 3.10.

License in practice

MIT license permits commercial and private use with minimal restrictions, making it suitable for most production deployments.

Quickstart

pip install langgraph-supervisor

from langgraph_supervisor import create_supervisor
from langgraph.prebuilt import create_react_agent

math_agent = create_react_agent(model=model, tools=[add, multiply], name="math_expert")
research_agent = create_react_agent(model=model, tools=[web_search], name="research_expert")

workflow = create_supervisor([research_agent, math_agent], model=model)
app = workflow.compile()
result = app.invoke({"messages": [{"role": "user", "content": "query"}]})

Requires Python >= 3.10. Requires langchain-core and langgraph to be installed and configured with a working LLM model.

Verify before relying

• Whether the library's supervisor pattern remains the recommended approach given the note recommending tool-calling directly for most use cases
• Performance characteristics when managing large numbers of agents or deeply nested hierarchies
• Compatibility guarantees with specific langgraph and langchain-core versions beyond the stated requirements

Package facts

Evidence: langgraph_supervisor-0.0.31-py3-none-any.whl

Tags

More Artificial Intelligence packages

Further reading