FlowAgent Now Built on LangChain and LangGraph

FlowAgent is built on LangChain and LangGraph. The latest version of FlowAgent, FlowAgent 2.0, was announced on November 10, 2025. LangChain quote-tweeted the launch the same day, calling out FlowAgent as a project using LangChain to orchestrate complex Web3 tasks.

This post explains what LangChain provides, what we built on top of it, and why we picked it as the orchestration layer underneath FlowAgent.

About LangChain

LangChain is an open-source framework for building applications with language models. LangGraph, the companion library, is a graph-based runtime for stateful, multi-step agent workflows. Together they give engineers a way to compose model calls, tool calls, retries, branching, and memory into one workflow definition that can be inspected, persisted, and replayed.

For an agent product like FlowAgent, the alternative is writing the orchestration logic from scratch on every workflow. LangChain and LangGraph let us avoid that work and focus on the Web3 surface area instead.

What we built on it

FlowAgent 2.0 ships several pieces that lean on LangChain and LangGraph directly:

• A visual workflow builder where each step in a multi-step Web3 action maps to a node in a LangGraph graph.
• Built-in persistence for in-flight workflows, so a long-running task can pause for a wallet signature and resume later without losing state.
• Cost-efficient execution, where the orchestration layer can choose smaller models for routing and confirmation steps and larger ones for steps that need stronger reasoning.
• A PDCA agent (Plan-Do-Check-Act) that plans a Web3 task, executes it, checks the on-chain result, and either confirms or retries. The PDCA loop is a LangGraph cycle by design.

Why we picked it

Two reasons drove the choice. The first is graph-shaped workflows. Web3 actions are not single calls. A simple stake-and-confirm action has at least four steps: read the user's balance, build the transaction, broadcast it, confirm the on-chain result. LangGraph's graph runtime is the right shape for that kind of work.

The second is the ecosystem. LangChain has a wide set of integrations, an active developer community, and tooling that engineers in our space already know. Standardizing on it shortens onboarding for new contributors and makes it easier for users who want to extend FlowAgent with their own nodes.

Source

The FlowAgent 2.0 launch and LangChain's quote-tweet are on X: @LangChain on FlowAgent (November 10, 2025). LangChain's project site is at langchain.com.