ant apply syncs Managed Agents from files in a repo

Anthropic has given its ant CLI an ant apply command, which keeps Claude Managed Agents resources on the API in sync with files in your own repository. The ClaudeDevs account on X announced it on September 3 and linked to the docs on the Claude platform site.

Five resource types can be declared this way: environments, agents, skills, memory stores and deployments. That is most of what a Managed Agents setup consists of. Anthropic launched the service in public beta on April 8: you define an agent’s tasks, tools and guardrails and run it on Anthropic’s infrastructure, where a built-in harness decides when to call tools, how to manage context and how to recover from errors. Infrastructure-as-code tools work the same way: the repository holds the desired state, and the tool changes the live resources to match it. With agent definitions stored as files, that setup finally has a history you can review.

A September 6 write-up from Get Claude Skills says ant apply creates and updates agents, environments, skills and deployments from those files, using a lockfile and a plan you approve before anything changes. That means a change to an agent’s setup can go through a pull request and code review before it reaches the API.