L

LangGraph

by LangChain, Inc.

Agent DevelopmentAI Agents & OrchestrationDeveloper Tools

Build agents as explicit state graphs — durable, interruptible and debuggable

Free · Subscription · Usage-based·Added Mar 19, 2026·Updated Aug 8, 2026
Share:
THE DAILY BRIEF
LangGraph

by LangChain, Inc.

Agent DevelopmentAI Agents & OrchestrationDeveloper Tools

Build agents as explicit state graphs — durable, interruptible and debuggable

Free · Subscription · Usage-based

LangGraph is an MIT-licensed agent runtime and low-level orchestration framework from LangChain for building stateful AI agents as explicit graphs of nodes and edges rather than opaque prompt chains. It gives engineering teams durable execution, human-in-the-loop checkpoints, cross-session memory and token-level streaming, and pairs with the commercial LangSmith platform for tracing, evaluation and deployment.

At a Glance

Category
Agent Development
Pricing
Free, Subscription, Usage-based
Target Market
CTOs, Platform Engineering Leaders, Enterprise Developers, Data Scientists, AI/ML Engineers
Deployment
Open-source, Self-hosted, Cloud-first, Hybrid
Founded
2023
Headquarters
San Francisco, California, United States
Customers
LangChain states it works with 35% of the Fortune 500

Key Features

  • Explicit graph state machine
  • Durable execution and checkpointing
  • Human-in-the-loop interrupts
  • Cross-session memory
  • Token-by-token streaming
  • MIT licence with no runtime lock-in

Capabilities

text generation
image generation
video generation
code generation
workflow automation
api access
audio generation
fine tuning
agent orchestration

Use Cases

  • Customer support agent with escalation
  • Production coding agent
  • Research and report generation
  • Regulated workflow automation
  • Multi-agent orchestration

Ideal For

Best For

  • Engineering teams building multi-step agents that must be debuggable and reproducible in production
  • Workflows requiring a human approval gate before an agent takes a consequential or irreversible action
  • Long-running agents that need durable state and the ability to resume after a crash or timeout
  • Teams already using LangSmith for tracing and evaluation who want deployment on the same platform
  • Applications that need token-level streaming of agent reasoning into an end-user interface

Not Ideal For

  • Simple single-call LLM features — the graph abstraction is pure overhead when there is no branching, state or retry logic to manage
  • Teams that prefer plain code: an HN discussion of frameworks versus hand-rolled agents argues that for experienced developers 'framework abstractions can add unnecessary complexity,' and that built-in parallelism 'can complicate debugging a lot'
  • Latency- or memory-constrained deployments where lighter runtimes make large efficiency claims — Agno's HN post claims 50x less memory than LangGraph
  • Small teams without the appetite to own graph maintenance, which the community itself flags as unsolved once you have several agents and shared state across subgraphs

Market Analysis

Developer-firstOpen-sourceEnterprise-grade

Pros

  • MIT-licensed and self-hostable, so the core framework carries no vendor lock-in and no runtime licence cost
  • The explicit state graph makes agent behaviour inspectable and reproducible, which is exactly what teams find missing in prompt-chain frameworks
  • Strongest named production reference list of any open agent framework — Klarna, LinkedIn, Uber, Cisco, Workday, Coinbase, Harvey, Bristol Myers Squibb
  • Actively maintained: 39,176 stars, 6,587 forks and commits pushed within a day of checking in August 2026

Cons

  • Steep learning curve — HN practitioners note that for experienced developers 'framework abstractions can add unnecessary complexity' compared with writing the agent loop directly
  • Built-in parallelism is a debugging hazard, described in an HN framework comparison as something that 'can complicate debugging a lot'
  • Maintainability at scale is an acknowledged community gap: a Show HN post observes that tutorials 'show you how to build a graph, not how to maintain one when you have 8 nodes, 3 agents, and shared state across subgraphs'
  • 673 open issues on the repository, and rival frameworks publish aggressive comparisons — Agno's HN post claims 5000x faster startup and 50x less memory
  • The free library funnels toward paid LangSmith, and LCU/LSU metering makes production cost hard to forecast before load testing

Pricing

LangGraph (open source)

$0

  • MIT licence
  • Self-host anywhere
  • No feature gating on the library itself

LangSmith Developer

$0 / seat per month

  • 1 seat
  • Up to 5,000 base traces per month, then pay-as-you-go
  • No managed deployment

LangSmith Plus

From $39/seat/mo

  • Unlimited seats
  • Up to 10,000 base traces per month, then pay-as-you-go
  • 1 free Serverless (Small) deployment

LangSmith Enterprise

Contact for pricing

  • Self-hosted and hybrid deployment
  • Custom SSO and RBAC/ABAC
  • Support SLA
  • Custom seats and workspaces

The LangGraph library itself is MIT-licensed and costs nothing — you pay only for LangSmith, LangChain's observability and deployment platform. Seats are $0 on Developer (1 seat, 5,000 base traces per month) and $39 per seat per month on Plus (unlimited seats, 10,000 traces, one free Serverless Small deployment), both then pay-as-you-go on overage. Managed compute and storage bill separately at $1.50 per LangChain Compute Unit and $1.00 per LangChain Storage Unit, which is where real production spend accumulates and what makes forecasting hard before load testing. Self-hosted and hybrid deployment, custom SSO, RBAC/ABAC and a support SLA are Enterprise-only and quoted.

Security & Compliance

soc2
gdpr
hipaa
iso27001
sso
data residency

THE DAILY BRIEF

Enterprise AI insights for technology and business leaders, twice weekly.

beri.net

Subscribe at beri.net/subscribe for twice-weekly AI insights delivered to your inbox.

LinkedIn: linkedin.com/in/rberi  |  X: x.com/rajeshberi

© 2026 Rajesh Beri. All rights reserved.

LangGraph is an MIT-licensed agent runtime and low-level orchestration framework from LangChain for building stateful AI agents as explicit graphs of nodes and edges rather than opaque prompt chains. It gives engineering teams durable execution, human-in-the-loop checkpoints, cross-session memory and token-level streaming, and pairs with the commercial LangSmith platform for tracing, evaluation and deployment.

LangGraph is an open-source agent runtime and low-level orchestration framework built by LangChain, Inc., the San Francisco company Harrison Chase and Ankush Gola founded in early 2023 and which now also has offices in New York, Boston and Amsterdam. Where a prompt chain hides control flow inside the model, LangGraph makes it explicit: developers define an agent as a graph of nodes and edges over a shared, typed state object, which yields deterministic branching, cycles, checkpointing, durable resume after failure, and human-in-the-loop interrupts at arbitrary points in a run. It ships persistent memory across sessions for what LangChain calls "rich, personalized interactions," native token-by-token streaming that surfaces agent reasoning and tool calls in real time, and moderation hooks intended to "prevent agents from veering off course." The library is MIT-licensed and free to use: 39,176 GitHub stars, 6,587 forks and 673 open issues as of August 2026, on a repository created in August 2023 and still receiving commits daily. Commercialisation happens next door in LangSmith — a Developer tier at $0 per seat with 5,000 base traces per month, Plus at $39 per seat with 10,000 traces and one free Serverless (Small) deployment, and Enterprise adding self-hosted and hybrid deployment, custom SSO and RBAC/ABAC, and a support SLA. Managed compute and storage bill separately at $1.50 per LangChain Compute Unit and $1.00 per LangChain Storage Unit. LangChain states it works with 35% of the Fortune 500, has crossed one billion open-source downloads, and ingests over one billion events per day on LangSmith; named LangGraph users include Klarna, LinkedIn, Uber, Cisco, Workday, Coinbase, Rippling, Lyft, Harvey, Abridge, Autodesk and Bristol Myers Squibb. It raised a $125M Series B led by IVP in October 2025 at a $1.25B valuation, bringing total funding to roughly $260M.

Ideal Buyer

Platform engineering teams putting a multi-step agent into production who need the run to be inspectable, resumable and pauseable for human approval — not a black-box prompt chain.

Key Benefit

Agent control flow becomes explicit code you can checkpoint, interrupt, replay and debug, instead of emergent behaviour you can only observe after the fact.

At a Glance

Category
Agent Development
Pricing
Free, Subscription, Usage-based
Target Market
CTOs, Platform Engineering Leaders, Enterprise Developers, Data Scientists, AI/ML Engineers
Deployment
Open-source, Self-hosted, Cloud-first, Hybrid
Founded
2023
Headquarters
San Francisco, California, United States
Customers
LangChain states it works with 35% of the Fortune 500

Key Features

  • Explicit graph state machine

    Agents are nodes and edges over a typed shared state object, making control flow inspectable rather than emergent from prompting.

  • Durable execution and checkpointing

    Runs persist their state so a crashed or paused agent resumes from the last checkpoint instead of restarting the whole task.

  • Human-in-the-loop interrupts

    Pause a graph at any node for approval or edit and then resume, which is essential before consequential or irreversible actions.

  • Cross-session memory

    Persists context between conversations so agents deliver personalised interactions across sessions rather than starting cold each time.

  • Token-by-token streaming

    Native streaming surfaces agent reasoning and tool calls in real time so users are not left watching a spinner.

  • MIT licence with no runtime lock-in

    The library is free and self-hostable anywhere; only LangSmith tracing, evaluation and managed deployment carry a cost.

Capabilities

text generation
image generation
video generation
code generation
workflow automation
api access
audio generation
fine tuning
agent orchestration

Use Cases

  • Customer support agent with escalation

    Route a ticket through classification, retrieval and drafting nodes, interrupting for human approval before any refund is issued.

  • Production coding agent

    Qodo publicly documented choosing LangGraph for its coding agent, citing the need for explicit control over multi-step code edits.

  • Research and report generation

    Fan out parallel retrieval nodes, merge results in shared state, and stream the synthesised report to the reader as it is written.

  • Regulated workflow automation

    Financial and healthcare teams use checkpoints and durable state to evidence exactly what an agent did at each step of a run.

  • Multi-agent orchestration

    Compose supervisor and worker subgraphs so specialised agents hand off work through one shared, inspectable state object.

Ideal For

Best For

  • Engineering teams building multi-step agents that must be debuggable and reproducible in production
  • Workflows requiring a human approval gate before an agent takes a consequential or irreversible action
  • Long-running agents that need durable state and the ability to resume after a crash or timeout
  • Teams already using LangSmith for tracing and evaluation who want deployment on the same platform
  • Applications that need token-level streaming of agent reasoning into an end-user interface

Not Ideal For

  • Simple single-call LLM features — the graph abstraction is pure overhead when there is no branching, state or retry logic to manage
  • Teams that prefer plain code: an HN discussion of frameworks versus hand-rolled agents argues that for experienced developers 'framework abstractions can add unnecessary complexity,' and that built-in parallelism 'can complicate debugging a lot'
  • Latency- or memory-constrained deployments where lighter runtimes make large efficiency claims — Agno's HN post claims 50x less memory than LangGraph
  • Small teams without the appetite to own graph maintenance, which the community itself flags as unsolved once you have several agents and shared state across subgraphs

Integrations

SDK Available
SDK:PythonJavaScriptTypeScript

Deployment

On-Premise

Market & Ratings

Estimated Customers

LangChain states it works with 35% of the Fortune 500

Market Analysis

Developer-firstOpen-sourceEnterprise-grade

Pros

  • MIT-licensed and self-hostable, so the core framework carries no vendor lock-in and no runtime licence cost
  • The explicit state graph makes agent behaviour inspectable and reproducible, which is exactly what teams find missing in prompt-chain frameworks
  • Strongest named production reference list of any open agent framework — Klarna, LinkedIn, Uber, Cisco, Workday, Coinbase, Harvey, Bristol Myers Squibb
  • Actively maintained: 39,176 stars, 6,587 forks and commits pushed within a day of checking in August 2026

Cons

  • Steep learning curve — HN practitioners note that for experienced developers 'framework abstractions can add unnecessary complexity' compared with writing the agent loop directly
  • Built-in parallelism is a debugging hazard, described in an HN framework comparison as something that 'can complicate debugging a lot'
  • Maintainability at scale is an acknowledged community gap: a Show HN post observes that tutorials 'show you how to build a graph, not how to maintain one when you have 8 nodes, 3 agents, and shared state across subgraphs'
  • 673 open issues on the repository, and rival frameworks publish aggressive comparisons — Agno's HN post claims 5000x faster startup and 50x less memory
  • The free library funnels toward paid LangSmith, and LCU/LSU metering makes production cost hard to forecast before load testing

Pricing

Free Trial Available

LangGraph (open source)

$0

  • MIT licence
  • Self-host anywhere
  • No feature gating on the library itself

LangSmith Developer

$0 / seat per month

  • 1 seat
  • Up to 5,000 base traces per month, then pay-as-you-go
  • No managed deployment

LangSmith Plus

From $39/seat/mo

  • Unlimited seats
  • Up to 10,000 base traces per month, then pay-as-you-go
  • 1 free Serverless (Small) deployment

LangSmith Enterprise

Contact for pricing

  • Self-hosted and hybrid deployment
  • Custom SSO and RBAC/ABAC
  • Support SLA
  • Custom seats and workspaces

The LangGraph library itself is MIT-licensed and costs nothing — you pay only for LangSmith, LangChain's observability and deployment platform. Seats are $0 on Developer (1 seat, 5,000 base traces per month) and $39 per seat per month on Plus (unlimited seats, 10,000 traces, one free Serverless Small deployment), both then pay-as-you-go on overage. Managed compute and storage bill separately at $1.50 per LangChain Compute Unit and $1.00 per LangChain Storage Unit, which is where real production spend accumulates and what makes forecasting hard before load testing. Self-hosted and hybrid deployment, custom SSO, RBAC/ABAC and a support SLA are Enterprise-only and quoted.

Security & Compliance

soc2
gdpr
hipaa
iso27001
sso
data residency

Connect

Sources

This page was written from 6 sources, 3 on domains other than langchain.com.

  1. 1.langchain.comlanggraphvendor
  2. 2.langchain.compricing langsmithvendor
  3. 3.langchain.comaboutvendor
  4. 4.github.comlanggraph
  5. 5.news.ycombinator.comitem
  6. 6.qodo.aiwhy we chose langgraph to build our coding agent
Newsletter

Stay Ahead of the Curve

Weekly enterprise AI insights for technology leaders. No spam, no vendor pitches—unsubscribe anytime.

Subscribe

Mentioned In

Enterprise AI

LangGraph vs Google ADK: Which Enterprise AI Framework Should You Choose?

Side-by-side comparison of LangChain's LangGraph and Google's Agent Development Kit for enterprise AI development. Real feature analysis, deployment costs, and use case recommendations for technical leaders and engineering leaders.

March 22, 2026
MCP

MCP vs LangChain vs OpenAI Functions: Which for Enterprise?

Choosing between MCP, LangChain Tools, and OpenAI Functions isn't an either/or decision—many teams use MCP for standardized data access alongside LangChain for orchestration. The real question is which to prioritize for your enterprise use case.

March 22, 2026
Geely

Geely Expands NVIDIA Partnership Across Physical, Enterprise, and Industrial AI

Geely Auto Group announced a major expansion of its NVIDIA partnership spanning autonomous vehicles, cloud-based agentic AI, and factory automation—signaling a transformation from car manufacturer to AI organization.

March 22, 2026
Cloud Infrastructure

Microsoft Loses OpenAI Exclusivity as AWS Pays $50B: What Enterprise AI Buyers Should Do Now

Microsoft Loses OpenAI Exclusivity as AWS Pays $50B. For enterprise decision-makers: strategic analysis, cost implications, and implementation guidance for A...

March 22, 2026
Enterprise AI

OpenAI Turns ChatGPT Into a Sales Channel for Retailers

OpenAI launched Agentic Commerce Protocol expansion for product discovery — Target, Walmart, and Shopify are integrated. High-intent shoppers meet conversational AI.

March 25, 2026
Enterprise AI

The $670K Gap: Why 78% of Enterprise AI Pilots Die Before Production—And How Governance Fixes It

Multi-agent AI systems promise 5x-10x returns, but 88% fail before launch. The missing ingredient isn't smarter models—it's operational infrastructure: orchestration, monitoring, and governance.

March 27, 2026

Latest Articles

View All →