Build with LangGraph or use a ready-made agent?
A ready-made agent is a product you sign up for and give a task. A framework such as LangGraph is code you write to build your own. They solve different problems, and most people asking this question want the first. This guide sets out what LangChain's own documentation says LangGraph is, what building one yourself involves, and when a ready-made agent is the shorter path.
What LangGraph is, per its documentation
LangChain's LangGraph overview calls it a low-level orchestration framework and runtime for building, managing and deploying long-running, stateful agents. It lists durable execution, streaming, human-in-the-loop and persistence as its focus, and says it lets you mix hand-coded deterministic steps with model-driven steps in the same graph. The same page says LangGraph is very low-level and recommends that people new to agents, or wanting a higher-level abstraction, start with LangChain's agents instead. It also names other parts of the LangChain family: Deep Agents (an agent harness), LangSmith (tracing, evaluation and deployment) and LangSmith Fleet (a no-code agent builder). These are the vendor's descriptions, and we have not tested any of them.
Build or use ready-made
| Question | Build with a framework | Use a ready-made agent |
|---|---|---|
| Who it suits | Developers who need control over each step and over where the model is used. | People who want a task done without writing code. |
| Time to first result | Longer: you design the graph, tools, prompts and hosting. | Shorter: sign up and describe the task. |
| Control and auditing | High. You decide the steps, state and approval points. Oversight is something you implement. | Limited to the controls the vendor offers. Check the permission and approval settings first. |
| Cost | The framework is open source. Model calls, hosting and any LangSmith services cost extra, and no complete workload price is published. | Published plan, credit or per-outcome prices, which vary by vendor. |
| Upkeep | You maintain it: updates, failures, safety checks. | The vendor maintains it, and can change features or prices. |
Choose a framework if
- You are a developer and the workflow has fixed steps that must run the same way every time.
- You need a human to approve specific actions at specific points, and the product options do not allow that.
- The agent must live inside your own systems and data.
Choose a ready-made agent if
- You want a result this week and do not write code.
- The job is common: personal admin, coding help, customer support or team workflows.
- You are happy with the vendor's data handling after reading it. Our checklist lists what to check.
Ready-made agents to look at first
- Poke, personal assistant. Price not verified.
- Claude Code, coding agent. From $20/month (Pro, per Anthropic pricing page); usage limits apply.
- ChatGPT workspace agents, workspace workflow agents. Workspace credit pools; no per-agent price claimed.
- Microsoft Copilot Studio, agent-building platform. $200 per pack per month for 25,000 credits; pay-as-you-go options.
- Fin, customer-facing service and sales agent. From $0.99 per outcome; seat fees and minimums may also apply.
See the LangGraph record, the use cases and what an AI agent is if you are starting out.
Spot an error? Tell us.