Building an AI agent that can interact with a blockchain network has never been a simple task. Between environment setup, boilerplate code, framework decisions, and testing infrastructure, the path from idea to a running agent has involved more friction than it should. The Hedera Agent Lab changes that.
Introducing Hedera Agent Lab: a fully integrated development workspace, accessible directly from the Hedera Developer Portal. Built on the open-source Hedera Agent Kit, it takes a complex multi-step process for creating agentic AI and simplifies it for developers so that by the time setup is finished, users have a working, chatting, transaction-executing agent ready to run against Hedera Testnet.
With the developer portal, users have a choice of using the no-code Agent Builder to instantly create a Hedera-powered agent using set templates, a low-code mode where users can choose their framework and plugins, then watch the agent source code built before their eyes, and an advanced mode where developers can customize their agent code. The platform offers the seamless sign in and wallet creation found with the other tools in the portal, and enables you to download your agent code so you can run it locally, or host your agent yourself.
The low-code Agent Lab in the Hedera Portal enables you to quickly choose plugins and frameworks to build a Hedera Agent
From zero to running agent in minutes
The no-code Agent Builder walks you through what kind of agent to build, which AI framework to use, and how your agent should execute transactions – either autonomously, or with a human-in-the-loop flow that requires a user signature before any on-chain transaction occurs.
The agent builder provides templated agents and a workflow to help you make informed decisions about configuring an agent.
Agent type
There are several pre-configured templates covering core use cases across the Hedera ecosystem – from a Basic agent for simple transactions and queries, through to an Everything agent with access to the full suite of Hedera tools. The next release of the Agent Lab will include templates with third-party and Hedera Stablecoin Studio plugins that enable you to add on additional functionality like swapping & trading tokens, borrowing and lending, and even creating your own stablecoins..
Framework
Users can choose between LangChain, Vercel AI SDK, and Google ADK(coming soon!) as the framework you use to build your agent. These frameworks help your agent to handle tools, execution, prompts & memory, and execution patterns. This selection will be reflected immediately in the generated code, making it easy to compare approaches, or stick with what you know.
Execution mode
For execution mode, you choose between Autonomous (the agent executes transactions automatically, ideal for testing) and Human-In-The-Loop (the agent surfaces unsigned transaction bytes for your review and approval before anything hits the network).
Once those three selections are made, the user is dropped into the full development workspace with a working agent scaffold already in the editor, ready to run.
In Human-in-the-Loop mode, users will be asked to sign all transactions before they are sent to Testnet
Build, run, and chat
The main workspace features a full in-browser code editor, a choice between ‘Block Mode’ with a configuration sidebar on the left with execution mode, and active Hedera plugins,for adjusting your framework or ‘Advanced Mode’ where you can edit code directly with the help of an AI assistant. There is also a terminal at the bottom that doubles as a live chat interface with your agent.
Click ‘Run Agent’ and your agent code executes directly in the browser. From that moment, users can type natural language messages – HBAR balance inquiries, account creation, message submission – and the agent responds with results.
An Open Chat button takes you into a full-screen conversation mode, and Advanced Mode exposes the complete underlying code alongside a built-in Chat Assistant that can explain what the code is doing, right next to the code itself.
Chat mode enables you to experience your agent the same way and end-user would
Coming Soon: A better way to build AI agents
The next features slated for development in the Hedera Agent Kit are the inclusion of the Google ADK as a framework option, and the addition of ‘Policies and Hooks’. The Google ADK will be integrated into the Agent Lab interface as soon as it is supported, and the Policies and Hooks, which are ways of adding controls and ‘hooks’ that enable you to trigger other events with the SDK, will be implemented in a ‘Controls’ panel on the left hand side.
The Controls tab will give you a no-code interface for setting hard behavioral limits on your agent. Cap maximum HBAR per transaction, restrict token interactions to an allow-list, enforce memos on all transactions, or prevent the creation of tokens with unlimited supply. These controls are backed by the Hedera Agent Kit’s Policies and Hooks system – deterministic, code-level enforcement that operates independently of the AI model.
Whether you are an experienced Web3 developer looking to add AI agent capabilities to your stack, a builder exploring the Hedera ecosystem for the first time, or an AI developer curious about what it looks like to give an agent real on-chain powers – the Hedera Agent Lab is built for you.
Get Started with the Hedera Agent Lab
Try Hedera Agent Lab: portal.hedera.com/agent-lab
Try the Agent Builder: portal.hedera.com/agent-lab#build
Explore the Hedera Agent Kit: https://github.com/hashgraph/hedera-agent-kit-js
Check out the Docs: https://docs.hedera.com/
Hedera AI Studio: https://hedera.com/ai-studio
Join the conversation: discord.gg/hedera
