Google’s Agents CLI, a tool for building and deploying AI agents on Google Cloud, has introduced a new Extension System. This system allows developers to use any agent framework—from ADK to LangChain and beyond—without being locked into a single framework or forking the CLI. For beginners building with AI agents, this means more flexibility and fewer barriers when choosing a framework for your project.
Why Flexibility Matters in Agent Frameworks
When Agents CLI launched, it was designed around the Agent Development Kit (ADK), Google’s proprietary framework. This worked well for teams using ADK, but the agent framework landscape has grown more diverse. Today, developers might choose LangChain, LangGraph, Deep Agents, or even custom in-house frameworks. Each framework has its strengths, and the best choice for a prototype might not be the best for production.
Enterprise teams also face governance requirements—like compliance scans or SSO gates—that need to be enforced across all projects. Previously, this meant either forking the CLI or building custom wrappers, both of which added complexity and maintenance overhead.
The new Extension System solves these problems by letting developers override or add commands to Agents CLI without forking the tool. This means you can use LangChain or another framework while still benefiting from Agents CLI’s deployment, evaluation, and observability features.
What Are Agents CLI Extensions?
An Agents CLI extension is a directory in a Git repository that contains a file called agents-cli-extension.yaml. This file declares which built-in commands the extension overrides and which new commands it adds. Extensions can:
- Override commands: Replace the behavior of built-in commands like
playground,deploy, oreval generate. - Add new commands: Introduce entirely new workflows as
agents-cli <your-command>.
Extensions are distributed through Git repositories, not a package registry. This design simplifies distribution and versioning, especially for enterprise teams that can host their extensions internally.
Example: LangChain Extension
Google has released a first-party LangChain extension that replaces ADK-specific commands with their LangChain equivalents. This lets developers scaffold, run, deploy, and evaluate LangChain agents through Agents CLI’s lifecycle. The extension is a project template that scaffolds a working LangChain agent wired into Agents CLI.
Here’s how it works:
- Scaffold the project: Run
agents-cli createwith the LangChain extension. This creates a project with no ADK dependencies. - Inspect the structure: The scaffolded project includes an
agents-cli-extension.yamlfile that carries the LangChain overrides. - Run and deploy: Use the same
agents-clicommands to run, deploy, and evaluate your LangChain agent.
Building Custom Extensions
The Extension System isn’t just for frameworks—it’s also for custom workflows. For example, an enterprise team might build an extension that wraps the deploy command with a Software Bill of Materials (SBOM) scan. This ensures every deployment meets compliance requirements without modifying the CLI itself.
Extensions can be referenced in several ways:
| Reference Type | Example |
|---|---|
| First-party shorthand | agents-cli extension add <name> |
| GitHub repository | agents-cli extension add acme/platform-extensions |
| Local path | agents-cli extension add local@../my-extension |
Extensions can be installed at two scopes:
- Project scope: The extension is committed to the project’s repository and picked up by teammates and CI pipelines.
- User scope: The extension is stored in the user’s home directory and applies to every project on the machine.
How Extensions Work Under the Hood
The Extension System follows the discipline of package managers like uv or npm. It resolves a Git reference to an exact commit SHA at install time, records the SHA in a lockfile, and uses the lockfile to re-materialize the exact code on every fresh checkout. This ensures consistency between local development and CI pipelines.
To advance a pin to the latest commit, developers run agents-cli extension update, which re-resolves the reference and re-prompts for third-party trust.
Frequently Asked Questions
Can I use multiple frameworks in one project?
No, the Extension System lets you choose one framework per project. However, you can scaffold multiple projects, each using a different framework.
Is the Extension System stable?
The Extension System is marked experimental in Agents CLI v1.8.0. The manifest schema and command surface may still change in breaking ways before the system is finalized. Teams depending on it should pin their Agents CLI version.
Can I use extensions with existing ADK projects?
Yes, extensions can be added to existing projects. However, overriding ADK-specific commands may break functionality if the project relies on them.
Conclusion
The Agents CLI Extension System is a significant step forward for developers building with AI agents. It removes the framework lock-in problem, letting you choose the right tool for the job without sacrificing the benefits of Agents CLI’s lifecycle management. Whether you’re using LangChain, Deep Agents, or a custom framework, extensions make it easier to scaffold, run, deploy, and evaluate your agents.
For beginners, this means less friction when experimenting with different frameworks and more confidence when moving from prototype to production. If you’re just starting out with AI agents, the LangChain first-party extension is a great place to begin.
Read next
- API Gateway adds streaming for real-time and AI APIs
- The Claude Code practicum on GitHub: 14 hands-on tasks with automatic checks
Want to try all of this hands-on? Start with the free Claude Code from Zero course.
Source
Based on Google Developers Blog’s announcement, “Powering the Agent Development Lifecycle with any framework, from ADK to LangChain and beyond”. Written for people learning to build with these tools.