Superstep vs LangSmith¶

Superstep and LangSmith both work with LangGraph agents, but they solve different problems and work well together. This page compares them on what each one is built for, so you can pick the right tool or combine them. Facts about LangSmith come from LangChain's own pages (listed under Sources) as checked on 24 October 2026; prices and plans can change, so check those pages before you decide.

In one sentence each¶

Superstep¶

• Visual first. Graphs are drawn on a canvas: nodes, edges, a typed state schema, loops, parallel branches and subgraphs. Editors do not need to write code; Code nodes are there when they want to.
• Self-hosted and open source. MIT license. Runs with Docker Compose or on a PaaS on your own infrastructure; your data stays there.
• Code export. Any workflow exports to a LangGraph Python or LangGraph.js file (see Export to Python).
• Human in the loop where people already are. Human nodes pause a run and ask for an answer in Slack, WhatsApp, the approvals inbox or a web chat (see Human approval and Chat).
• Vietnamese and English UI and handbook.
• Built-in run tools. Step-by-step runs, run history, workflow tests with checks (see Workflow tests) and OpenTelemetry traces.

LangSmith Studio¶

According to the Studio documentation, Studio is "a specialized agent IDE" for agents that implement the Agent Server API protocol, integrated with tracing, evaluation and prompt engineering. Its listed features include visualizing the graph, running and interacting with the agent, managing assistants and threads, iterating on prompts, running experiments over a dataset, managing long-term memory, debugging state with time travel and one-click deploy to LangSmith Cloud. It works with graphs deployed on LangSmith or running locally through the Agent Server (for example langgraph dev), and has a graph mode and a chat mode.

Where LangSmith is strong:

• Tracing and evaluation depth. LangSmith is LangChain's observability and evaluation platform; datasets, experiments, feedback and prompt iteration are first-class.
• Managed deployment. The Plus plan includes LangSmith Deployment (one free small serverless deployment), with more capacity charged on usage.
• Code-first control. Your graph is your own Python or JavaScript code.

LangSmith Agent Builder¶

LangChain announced Agent Builder as generally available on 13 January 2026. It is a no-code product: you describe the goal in your own words, and it writes the instructions, picks tools and adds subagents when needed. Agents learn from feedback through memory, can ask permission before sensitive actions, can be shared and cloned in a workspace, connect tools through a remote MCP server and can be called through an API. The announcement says it is available on all LangSmith plans. On the current pricing page, the no-code product for building agents "using everyday language, not code" is listed as LangSmith Fleet.

Agent Builder and Superstep are both ways to build agents without writing code, with different approaches: Agent Builder lets the agent plan its own steps from a goal; Superstep has you draw the steps, so every path, branch and approval is explicit and reviewable. Where the model should choose, a Superstep Agent node picks and calls the tools you list inside one drawn step, and asks for approval before the tools you mark.

Pricing and hosting¶

The self-hosted LangSmith overview states that "Self-hosted deployments require an Enterprise plan and the LangSmith license key delivered with that plan", and offers three topologies: full platform, hybrid and a standalone Agent Server.

Using them together¶

Superstep speaks LangGraph, so you do not have to choose one side:

A common setup: business users and reviewers build and approve workflows in Superstep, developers export them as LangGraph projects when they need custom code, and the team uses LangSmith for tracing and evaluation across both.

Which one to pick¶

• Choose Superstep when you need self-hosting without an Enterprise contract, a visual editor for non-developers, approvals in Slack, WhatsApp or a web chat, a Vietnamese UI, or an open-source license.
• Choose LangSmith Studio when your agents are written in code and you want LangChain's tracing, evaluation and managed deployment in one place.
• Choose LangSmith Agent Builder when you want agents that plan their own steps from a plain-language goal, hosted by LangChain.

Sources¶

• LangSmith Studio
• LangSmith pricing
• Deploy to self-hosted
• Now GA: LangSmith Agent Builder