LangGraph Platform
San Francisco, USA
A low-level orchestration framework for stateful agents, with a managed service for deploying them in production.
What it is
LangGraph is a low-level, MIT-licensed Python/JS framework (from LangChain) for building stateful, graph-structured agent workflows. The managed side — now LangSmith Deployment, formerly branded LangGraph Platform — runs those agents with durable execution, versioning, and rollbacks.
At a glance
7/7 facts sourced- Pricing
- Framework is free (MIT). Managed: Developer $0/seat (up to 5k base traces/mo), Plus $39/seat/mo, Enterprise custom; usage add-ons include deployment runs at $0.005/run.source ↗
- Hosting model
- Open-source framework self-hosted on your own infra; managed cloud deployment available; self-hosted and hybrid deployment gated to the Enterprise tier.source ↗
- Memory & state
- Durable execution across restarts and cold starts, persistent checkpoints (payloads up to 25 MB), and semantic search for long-term memory, with per-subagent state isolation via threads.source ↗
- MCP support
- Nativesource ↗
- Integrations
- Docs cite 1000+ integrations; tools are consumed as built-ins plus remote (HTTP) and local MCP servers via langchain-mcp-adapters.source ↗
- Model providers
- Many via langchain-<provider> packages: OpenAI, Anthropic, Google (Vertex + GenAI), AWS, Ollama, Groq, Azure AI, Mistral, xAI, Cohere, DeepSeek, and more.source ↗
- Open source
- Yessource ↗
Editorial scorecard
Overall 4.3 / 5
Our opinion — the same 0–5 rubric is applied to every platform, including Alfe. See the rubric.
Strengths
- Core framework is open source (MIT) and self-hostable.
- Broad model-provider coverage through a standard interface.
- Managed tier offers durable execution, persistent checkpoints, and human-in-the-loop.
Trade-offs
- Managed usage is metered across many dimensions (traces, runs, uptime, LCUs), making cost harder to predict.
- Self-hosted and hybrid managed deployment are gated to the Enterprise tier.
- In the OSS framework, MCP is a separate adapter library rather than built into the core.
Our take
Editorial opinion — not a hands-on test resultIn our opinion, LangGraph is a strong choice when you want low-level control over stateful, multi-step agents and are comfortable with a graph mental model. The managed service adds the production plumbing — durable execution, checkpoints, rollbacks — that teams otherwise have to build. The main trade-off is cost predictability: the metered pricing has many dimensions, and running fully on your own infra pushes you toward Enterprise.
Sponsored
Alfe is a managed agent-hosting platform. If you want an agent hosted and running without wiring infrastructure yourself, see Alfe → (or read its dossier).
FAQ
- Is LangGraph free to use?
- The framework is MIT-licensed and free. The managed service has a $0 Developer tier plus pay-as-you-go usage, with a $39/seat/mo Plus tier.
- Can I run LangGraph on my own infrastructure?
- Yes — the framework is a self-hosted library. Self-hosted or hybrid managed deployment is available on the Enterprise tier.
Compare LangGraph Platform
Sources
- LangSmith Deployment (LangGraph Platform) · accessed 2026-07-01
- LangChain pricing · accessed 2026-07-01
- LangGraph MCP docs · accessed 2026-07-01
- LangGraph on GitHub (MIT license) · accessed 2026-07-01
The Agent Examiner editorial · last updated 2026-07-01. Facts are drawn from each vendor’s own public materials; scores are our editorial opinion (methodology).
Product and company names are trademarks of their respective owners and are used here only to identify the products under discussion.