Le Répertoire des AgentsFR · EN · ES · JA · HI
Home › Advanced agents

“I want a custom agent workflow, with every step under control”

Developers who find black boxes unbearable: you want to decide exactly what happens at each step, where the agent may loop, and where a human must approve before continuing.

Recommended agent: LangGraph
💰 Free open-source library; you pay for the model calls, and for LangSmith if you take it (free to start, then $39/user/month). Level: Advanced 🌍 United States Free open-source library (MIT license), with no time limit; associated hosted services (LangSmith) optional

Price verified on 2026-10-29.

LangGraph (from LangChain, langchain.com), an open-source library that models your agent as a graph: nodes (steps, model calls, tools), edges (transitions, conditions, loops) and a shared state. It's the tool for serious agent workflows, with error recovery and human approval at the points you choose.

How it works

Trigger: An event starts the graph's execution — flows the state from step to step, calls models and tools, loops if needed and stops at the planned approvals, and you get an agent process that's deterministic where it needs to be, autonomous where it's safe.

Step-by-step setup

  1. Install LangGraph in your Python or JavaScript project and plug in your model's API key.
  2. Draw your graph: the steps, the transition conditions and the human approval points.
  3. Add persistence so the run resumes after a pause or an error.
  4. Test every branch with edge cases before connecting real users.
⚠️ Guardrails: A graph can loop: set a maximum number of iterations per run and a budget per task. Plan human approvals before any irreversible step (sending, payment, deletion).
🔧 Technical details
Installation
API / SDK
Prerequisites
Python (pip install -U langgraph) or JavaScript/TypeScript (LangGraph.js); API keys for the model providers integrated into the application.
Hardware
non_applicable
Self-hosting
yes — open-source library run in your infrastructure; production deployment possible via LangSmith Deployment (LangChain platform, cloud and self-hosted options).
License
open_source — MIT License (MIT), LICENSE file in the official langchain-ai/langgraph repo.
API & integrations
API: yes — deployed LangGraph agents expose an API via the LangGraph server; Python and JS API reference published. · SDK: yes — official Python SDK and JavaScript/TypeScript SDK. · CLI: yes — LangGraph CLI (langgraph-cli) documented for local development and deployment. · Webhooks: Not documented · MCP: yes — MCP tools usable via the LangChain ecosystem's official adapter (langchain-mcp-adapters).
Models
Model-agnostic: LangGraph orchestrates models via LangChain or any other client; the LangChain ecosystem's model integrations accept your own keys (BYOK). Local models possible via compatible integrations (e.g. Ollama in the LangChain ecosystem).
Data
Agent state and memory are persisted by the application via checkpointers and a store in the infrastructure chosen by the developer; tracing to LangSmith (observability) is an optional add-on.
Operations
Persistence and recovery after failure (durable execution), human-in-the-loop, short- and long-term memory; debugging and tracing via LangSmith (execution paths, state transitions, run metrics), according to the official documentation.

Not documented in the official sources: Documented limits.

Official sources: docs.langchain.com · github.com — verified on November 1, 2026.

Find your agent with the quiz →

Frequently asked questions

How much does LangGraph cost?

Free open source library; you pay for model calls, and for LangSmith if you take it (free to start, then $39/user/month). Free open source library (MIT license), with no time limit; associated hosted services (LangSmith) optional. Price verified on October 29, 2026.

Is LangGraph free?

Yes. Free open source library (MIT license), with no time limit; associated hosted services (LangSmith) optional.

Is LangGraph hard to get started with?

Level: Advanced. Setup takes 4 steps; the first: Install LangGraph in your Python or JavaScript project and plug in your model's API key.

What limits and guardrails should I know about with LangGraph?

A graph can loop: set a maximum number of iterations per run and a budget per task. Plan human approvals before any irreversible step (sending, payment, deletion).

Other agents for this problem

LLM Council

I want a solid second opinion before a big decision

Free software (public repository, no open-source license); you pay for the model calls (about 2 to 3 times the cost of asking a single model).

CrewAI

I want a team of agents that splits up a big job

$0 for the open-source framework and the basic platform (50 runs/month); enterprise platform by quote, plus the cost of the model calls.

AutoGen

I want agents that debate each other until they find the solution

Free and open source; you pay for the model calls used by the agents — a multi-round debate costs several times a simple answer.

MCP (Model Context Protocol)

I want to plug my own tools and data into my agents

Free: it's an open standard, not a product; each tool you plug in keeps its own pricing.

Ollama (agents locaux)

I want agents that run at home, without sending my data to the cloud

$0: the software and the open models are free; the real cost is your computer (a good graphics card makes the models fast and pleasant).

See also

“I want a solid second opinion before a big decision” — LLM Council“I want a team of agents that splits up a big job” — CrewAI“I want agents that debate each other until they find the solution” — AutoGen

Profile updated on November 1, 2026 — see what's new at the Directory.