LangGraph

Summary

LangGraph is LangChain's low-level orchestration framework and runtime for building long-running, stateful agents as graphs of deterministic and LLM-driven steps. It supplies durable execution, persistence, streaming and human-in-the-loop interrupts, and deploys through LangSmith.

Best for: Engineering teams building production agents who need cyclical graphs, durable state and replayable runs
Not for: Teams that want to ship a first working agent in under an hour, do not need graph-level control over state and branching, or are building a lightweight TypeScript app on serverless infrastructure where a framework like Mastra fits the deploy target better.

Captured 2026-09-11 · theagentsindex.com

From: Freethen $39/month
Free tier: YesLangGraph itself is MIT-licensed and free to use. The LangSmith Developer plan is $0 for a maximum of 1 seat, with 5k base traces/month included then pay-as-you-go, 5 LCU of Fleet usage and 5 LCU plus 1 LSU of Sandbox usage per month; the Deployment row reads N/A for Developer, so managed deployment requires Plus.

Visit LangGraph Compare Find alternatives

What it does

Debugs and fixes: PartlyLangGraph itself does not diagnose code; the separate LangSmith Engine product analyses LangGraph agent traces, proposes fixes and can open a pull request with one.LangSmith Engine detects issues in your LangGraph agent traces and proposes fixes. You can open a pull request with the proposed fix directly from…docs.langchain.com/oss/python/langgraph…
Done-for-you service: NoYou run the software. On selected annual contracts LangChain assigns an AI engineer, but the published scope is advisory - best practices, guidelines, troubleshooting and reviews.LangChain AI Engineers work in close collaboration with the Customer to impart best practices, establish guidelines, facilitate troubleshooting, and…langchain.com/support-plans

How much it does unattended

Runs autonomously: YesBuilt for unattended long-running work: checkpointed durable execution lets a run survive failure and resume, and human-in-the-loop checks are described as something you add rather than a default.Persistence: Build agents that persist through failures and can run for extended periods, resuming from where they left off.docs.langchain.com/oss/python/langgraph…
Multi-agent: YesSingle, multi-agent and hierarchical control flows are all expressed in the same framework; the LangChain docs enumerate subagent, handoff, router and custom-workflow patterns.Design diverse control flows — single, multi-agent, hierarchical — all using one framework.langchain.com/langgraph
Agent permissions: PartlyThere are no preset permission tiers: you decide where the agent must stop by placing interrupt() calls, which pause the graph and wait for approval before continuing.Add human-in-the-loop checks to steer and approve agent actions.langchain.com/langgraph

Which models it runs on

Claude: YesClaude models are reached through the first-party ChatAnthropic integration, and also via ChatBedrock and Vertex for Anthropic models on those clouds.You can find information about Anthropic's latest models, their costs, context windows, and supported input types in the Claude docs.docs.langchain.com/oss/python/integrati…
GPT: YesFirst-party ChatOpenAI integration, plus AzureChatOpenAI for GPT models hosted on Azure; both are listed in the featured chat-model table.You can find information about OpenAI's latest models, their costs, context windows, and supported input types in the OpenAI Platform docs.docs.langchain.com/oss/python/integrati…
Open models: YesOpen-weight models run locally through the ChatOllama integration, and self-hosted vLLM and Hugging Face endpoints are documented as chat-model integrations.Ollama allows you to run open-source Large Language Models (LLMs), such as gpt-oss, locally.docs.langchain.com/oss/python/integrati…
Model choice: YesYou choose the model. LangGraph does no model abstraction of its own and the vendor markets provider-independence; the featured integration table spans OpenAI, Anthropic, Google, Bedrock, Ollama and more.Build and ship agents fast with any model provider. Use high-level abstractions or fine-grained control as needed.langchain.com/langgraph

Where you use it

In your editor: PartlyNo editor extension is documented. The vendor's own development surface is LangSmith Studio, a browser IDE that attaches to a locally running Agent Server; your own editor is where you write the graph.Studio: A specialized IDE for visualization, interaction, and debugging. Connects to a local Agent Server for developing and testing your graph.docs.langchain.com/langsmith/components
On the command line: Yeslanggraph-cli runs the Agent Server locally with langgraph dev, and builds, dockerises and deploys with build, dockerfile and deploy.LangGraph CLI is a command-line tool for building and running the Agent Server locally.docs.langchain.com/langsmith/cli
In your pipeline: YesThe vendor publishes a CI/CD reference pipeline with a companion repository: tests and offline evals on a PR, then quality-gated preview and production deploys via the Control Plane API.This guide demonstrates how to implement a comprehensive CI/CD pipeline for AI agent applications deployed in LangSmith Deployment.docs.langchain.com/langsmith/cicd-pipel…
In a browser: YesDeployments are created and managed from the LangSmith web control plane at smith.langchain.com, and Studio runs in the browser.Control Plane: The UI and APIs for creating, updating, and managing Agent Server deployments.docs.langchain.com/langsmith/components

Whose machine it runs on

Self-hosted: YesThe library is MIT-licensed and installed with pip or npm, so it runs wherever you install it; self-hosting the managed Agent Server platform is a separate Enterprise hosting option.Yes. LangGraph is an MIT-licensed open-source library and is free to use.langchain.com/langgraph
Open source: YesMIT-licensed library developed in the open by LangChain Inc; the product page's primary call to action links to the GitHub repository.Yes. LangGraph is an MIT-licensed open-source library and is free to use.langchain.com/langgraph

What it costs to run

How it meters: YesTwo meters at once on the managed platform: a per-seat licence at 39 dollars per seat per month, plus consumption in LangChain Compute Units (1.50 dollars) and Storage Units (1.00 dollar). The library itself is free.LangChain Compute Unit (LCU): A LCU is a normalized unit of work done on the LangSmith Platform and is used for pricing purposes.langchain.com/pricing
Free tier: YesNo call needed: the framework is free under MIT, and the LangSmith Developer plan is 0 dollars per seat with 5k base traces a month for self-serve signup.Yes. LangGraph is an MIT-licensed open-source library and is free to use.langchain.com/langgraph
API access: YesThe Agent Server is an API: runs, threads, assistants, state and cron endpoints, available locally from the CLI at no cost. On the managed platform the Assistants API and the 30-plus endpoint set are Plus and Enterprise rows.The resulting server exposes all API endpoints for runs, threads, assistants, etc., and includes supporting services such as a managed database for…docs.langchain.com/langsmith/cli
MCP server: YesA deployed agent can be exposed as an MCP server, so another agent can call it; the plan table checks this row on Developer, Plus and Enterprise alike.Expose agent as MCP serverlangchain.com/pricing
Bring your own key: YesYou supply provider keys directly in your own code, and the LLM Gateway's bring-your-own-key row is checked on every plan. Enterprise adds hybrid and self-hosted so inference and data stay in your cloud.BYO model API keylangchain.com/pricing

Buying it for a team

A company can buy it: YesSold to organisations with workspaces and per-seat billing; Enterprise adds annual invoicing, custom terms and an infosec review.A seat is a distinct user inside your organization. We consider the total number of users (including invited users) to determine the number of seats…langchain.com/pricing
Seat model: YesFloor of one seat, no ceiling: Developer is capped at a single free seat, Plus adds unlimited seats at 39 dollars each per month, Enterprise is custom.Add unlimited seats $39 per seat/monthlangchain.com/pricing
Pooled budget: YesThe included usage allowance is one organisation-wide balance that resets monthly, not a per-seat grant: 5 LCU on Developer, 25 LCU on Plus.Developer & Plus plans include a free usage allowance, shared across your organization and resets monthlylangchain.com/pricing
Admin controls: PartlyWorkspace RBAC and tag-based ABAC are Enterprise-only; other plans default every user to Admin. Gateway cost controls, rate limiting and PII redaction are checked on all three plans.RBAC (Role-Based Access Control) is an Enterprise feature for managing workspace-level permissions.docs.langchain.com/langsmith/rbac
Audit log: PartlyAdministrative action is recorded, but only on Enterprise: a tamper-resistant log of settings, membership, credential and workspace changes, retained up to 400 days and queryable by API.Audit logs are available on Enterprise plans. If you're interested in upgrading to Enterprise, contact our sales team.docs.langchain.com/langsmith/audit-logs
Single sign-on: PartlySAML SSO and SCIM are Enterprise-only; Plus gets Google and GitHub social login and Developer none. Self-hosted installs can wire almost any OIDC provider.Enterprise customers can configure SAML SSO and SCIM.docs.langchain.com/langsmith/authentica…

What happens to your code

Opt out of training: YesNothing to opt out of: the vendor states it does not train on customer data. Separately, setting LANGSMITH_TRACING=false keeps agent data on your own machine entirely.No. LangSmith does not use your data to train models. Your traces, prompts, and outputs remain private to your organization.langchain.com/pricing
Data residency: PartlyUS or EU choice inside LangChain's cloud from Plus upward; Enterprise adds hybrid and self-hosted with the data plane in your own VPC. The Developer column reads N/A.LangChain's Cloud (US or EU)langchain.com/pricing
Getting out: PartlyBulk trace export to an S3-compatible bucket in Parquet is Enterprise-only for anyone who signed up after 3 August 2026. Usage is metered in arrears rather than prepaid, so there is no credit balance to expire.For customers who signed up after August 3, 2026, bulk export is only available on the LangSmith Enterprise plan.docs.langchain.com/langsmith/data-export
Certifications: YesSOC 2 Type II, GDPR and HIPAA, with annual third-party audits and penetration testing; data encrypted AES-256 at rest and TLS 1.2 or higher in transit.We maintain SOC 2 Type II, GDPR, and HIPAA compliance and undergo annual third-party audits and penetration testing.docs.langchain.com/langsmith/shared-res…

30 sourced claims on this page. Checked 2026-09-02. How we check.

Alternatives

Where we would send a reader instead, in that listing's own words.

LangChain
Open source agent frameworks plus LangSmith, a hosted platform to trace, evaluate, deploy and govern agents.
CrewAI
Open-source multi-agent framework with a managed platform for deploying and governing agent workflows.
Mastra
Open-source TypeScript framework for building AI agents, with a hosted platform for running and observing them.
Pydantic AI
Open-source Python agent framework from the Pydantic team, with a typed agent loop and swappable models.

Filed under

APIAutonomousClaudeEnterpriseFree tierGPTMulti-agentOpen modelsOpen sourceSelf-hosted

Sources and updates

The pages we read21

1 of 21 independent of the vendor.

Platform record2026-08-30

Licence: MIT
Stars: 40,706
Forks: 6,863
Last push: 2026-08-30

Updates4

2026-09-02Capabilities · Commercial terms · Free tier · Price
2026-08-31Capabilities · Commercial terms · Free tier
2026-08-23Alternatives · Documentation link · One-liner · Repository · Summary
2026-08-16One-liner · Summary