Hedera Agent Lab: No-Code AI Bots Now Come With Training Wheels

Hedera just dropped Hedera Agent Lab, a browser-based development platform that lets developers build on-chain AI agents using no-code, low-code, or full custom code workflows—all accessible from the Hedera Developer Portal. Building AI agents on Hedera just got easier, and by "easier" we mean "your grandma could probably do it if she squinted at the screen hard enough."

Agent Lab is now live on the Developer Portal—a fully integrated, browser-based environment to build, run, and chat with your Hedera agent in minutes. No-code. Low-code. Full control. You pick. It's like a vending machine, except instead of snacks, you get blockchain-interacting AI bots, and instead of coins, you pay with your sanity trying to understand what "agentic" actually means.

The platform is built on the open-source Hedera Agent Kit and supports LangChain and Vercel AI SDK as framework options, with Google ADK support planned for a future release. It's designed to reduce the setup friction developers typically face when connecting AI agents to a live blockchain network—because nothing says "fun Friday afternoon" like manually configuring environment variables until 2 AM.

How It Works

Hedera describes the traditional process of building an AI agent capable of interacting with a blockchain as involving multiple friction points: environment configuration, boilerplate code, framework selection, and testing infrastructure. Agent Lab consolidates all of these steps into a single, integrated workspace inside the developer portal. Think of it as the difference between assembling IKEA furniture with tools from a hardware store versus just clicking it together like LEGO. Both get you a shelf. One doesn't require therapy afterward.

Once a developer completes the initial setup choices, they receive a working agent scaffold already loaded in an in-browser code editor, connected to the Hedera Testnet and ready to execute. It's basically a "here's your starter pack" moment, except instead of Pokémon, you're collecting on-chain transactions.

Three Development Paths

No-code Agent Builder: Uses pre-configured templates to generate a working agent without writing any code. Templates range from a Basic agent for simple queries and transfers to an Everything agent that accesses the full suite of Hedera tools. This is the "easy mode" for people who think console commands are scary and debugging is a personal attack.

Low-code mode: Lets developers select their preferred AI framework and plugins. The agent source code is generated in real time based on those selections. It's the middle ground between "I want to click buttons" and "I got C++ at 3 AM energy."

Advanced mode: Provides direct access to the full underlying code, alongside a built-in AI assistant that explains what the code is doing as developers edit it. For the "I learned to code from Stack Overflow threads and trauma" crowd.

The platform also includes seamless wallet creation and sign-in, and developers can download their agent code to run locally or self-host. Take it home, put it in your garage, teach it tricks. Just don't name it.

Framework and Execution Options

Developers building on Agent Lab choose from three configuration options before entering the main workspace. Framework selection currently includes LangChain, Vercel AI SDK, and a coming-soon option for Google ADK. These frameworks manage how the agent handles tools, memory, prompts, and execution patterns. Switching between them updates the generated code immediately, making it straightforward to compare approaches. It's like test-driving different cars, except the cars are abstract software paradigms and you're still paying for gas in the form of transaction fees.

Execution mode determines how the agent handles on-chain transactions. Two options are available:

Autonomous mode: The agent executes transactions automatically. Useful for development and testing scenarios. Let it cook. Let it burn. Let it learn.

Human-in-the-the-Loop mode: The agent surfaces unsigned transaction bytes for the developer to review and approve before anything is submitted to the network. This is relevant for production use cases where transaction oversight matters. The responsible adult mode. The "please don't liquidate my entire portfolio because a prompt injection told you to" setting.

Once these selections are made, developers enter the workspace and can interact with their agent through a terminal that doubles as a live chat interface. Typing natural language commands, such as checking an $HBAR balance, creating an account, or submitting a message, returns real results from the Testnet. It's like having a very obedient digital intern who never sleeps and only works in testnet conditions. Dream employee, honestly.

Coming Features: Policies, Hooks, and Stablecoin Studio

Hedera has outlined two major additions planned for upcoming releases of Agent Lab. The first is integration with the Hedera Stablecoin Studio plugin, which will enable agents to perform token swaps, lending, borrowing, and stablecoin creation directly from within the platform. Your agent, now with DeFi powers. Handle with caution and maybe a helmet.

The second is a Controls panel built on the Agent Kit's Policies and Hooks system. This gives developers a no-code interface for setting hard behavioral limits on their agents. Examples include:

• Capping the maximum $HBAR amount per transaction
• Restricting token interactions to an approved allowlist
• Requiring memos on all outgoing transactions
• Blocking the creation of tokens with unlimited supply

These controls operate at the code level, independently of the AI model itself. That distinction matters for developers building agents that need predictable, auditable behavior in production environments.