Ably

Documentation

npx -p @ably/cli ably init now installs the CLI globally as part of the onboarding step, so the Ably CLI is available in every shell from then on. Run ably init --help to list available flags for finer control. The install step is non-blocking: if it can't complete, for example because npm install -g would require elevated permissions on a system-wide Node install, init carries on with authentication and Agent Skills installation, and tells the user how to install manually afterwards.

The headline addition is ably init, a single command that takes a new developer from zero to fully set up. Run it through npx @ably/cli ably init and it will install the CLI on your machine, walk you through the OAuth login flow, and install the Ably Agent Skills bundle into whichever AI coding tools you have configured, all in one step. No separate npm install -g, no manual auth, no copy-pasting skill files into your editor's config directory. One command, and you're ready to build with Ably from your terminal or from inside Claude, Cursor, Windsurf, or VS Code. The skills themselves are also installable on their own via ably skills install, for developers who already have the CLI set up and just want to add or refresh the AI assistant integration. Out of the box the CLI supports Claude Code, Cursor, VS Code, and Windsurf, with auto-detection of whichever of these you already have on your machine. Run ably init --help or ably skills install --help to see the available flags for finer control.

Version 1.0.0 of the Ably CLI has been released, marking the first stable major version of @ably/cli. For anyone wiring the CLI into automation or AI agents, long-running commands now emit a structured JSON envelope with status and completed signals. Agents can tell exactly when a connection is live, a stream is flowing, or a command has finished. No more scraping log lines or guessing at exit timing. When a command fails, errors include a hint field that suggests the next step. The hint usually names a specific command to run resulting in agents that can self-heal in fewer turns, without a human in the loop.