Anthropic announced on X a new ant CLI for Claude Platform that lets every API endpoint run from the terminal, including the Messages API and Claude Managed Agents. The account states that the tool can also pipe results straight into the shell and is understood by coding agents through the claude-api skill.
The company’s post describes an interactive login flow available through ant auth login. According to the account, that command runs a browser OAuth flow, scopes the token to a workspace, stores it, and then reuses the same credential for CLI and SDK API calls. A promotional screenshot included with the post shows the login flow, alongside copy claiming there are “no keys to copy, paste, or rotate.”
ClaudeDevs also claims that every API resource is exposed as a subcommand, including messages, models, and files. The CLI reportedly supports building requests from typed flags or piped YAML, inline file references with @.path, and response reshaping with --transform. Output options listed in the post include json, yaml, jsonl, and explore, the latter described as a TUI for folding and searching results.
Managed Agents are another focus of the release. The post says the CLI can create and update agents locally or from CI, with agents versioned as YAML in Git and synchronized using ant beta:agents update. It also says a user can start a Managed Agents session, send an event, and later retrieve a trace covering “every event, every tool call, every decision.”
Claude Code is also folded into the workflow, according to the post. ClaudeDevs says it can use ant through the built-in /claude-api skill to list sessions, upload a folder of PDFs, or debug a run, with installation available through brew, curl, or go.
Source: ClaudeDevs on X
