Google says Agents CLI now supports extensions, which can override built-in commands or add new ones. The system arrived in version 1.5.0, and version 1.8.0 adds a first-party LangChain extension. An extension is a directory in a Git repository with an agents-cli-extension.yaml manifest, installed at project or user scope.
The LangChain template scaffolds an agent with no google-adk dependency and deploys to Cloud Run through the same path as ADK projects. It overrides six commands, including run, playground and eval generate. It refuses eval dataset and eval optimize because they need an ADK agent object, while deploy, infra and login stay unchanged.
The post also builds a sample sbom-deploy extension that runs a software bill of materials scan before calling the native deploy command. Manifest run entries are command vectors rather than shell commands, so chaining happens in a wrapper script. An environment variable set by the loader prevents the wrapper from calling its own override. The system is marked experimental in v1.8.0.
Why it matters if you build agents
Teams can use Agents CLI deployment, evaluation and observability commands with LangChain or LangGraph agents without adopting ADK.
Platform teams can enforce checks such as scans or login gates across projects by publishing an internal extension repository.
Because the manifest schema and command surface may change, pin the Agents CLI version and the extension's commit SHA in the lockfile.
Get stories like this every weekday
One short email with the day's agent tooling news.