17 July 2026Changelog

Changelog #0005

This last week for us was all about enhancing the Cleotic CLI. Full read-write capability, scriptability improvements, and features that provide a more efficient experience for AI agents.

Avatar for Andy Bright

Andy Bright

Director of Product and Technology

Featured Image for Changelog #0005

Happy Friday friends. This last week for us was all about the Cleotic CLI. Opening up full read-write across all product capabilities and making it a first-class citizen for scripts, CI pipelines, and agents. Here is the rundown.

CLI now fully scriptable

If you drive Cleotic from a terminal, a cron job, or an AI agent, the CLI now holds a firm machine contract across every command:

  • Every read and write command supports --json, and mutating commands support --no-input. When you ask for JSON you get exactly one JSON document on stdout (progress and prompts stay on stderr), so piping into jq (which agent's just love to do) just works.

  • Exit codes are stable and meaningful: 2 means you got the usage wrong (with guidance pointing at the flag you missed), 1 means the operation failed, 0 means it worked.

  • Deletes are safe by default. Destructive commands ask for confirmation in a terminal, require an explicit --yes when running non-interactively.

The few commands that previously only spoke human - projects use, config set, config get, and config reset - now speak JSON too.

No HITM UI work required for new orgs

Previously a brand-new organization needed someone to sign in through the browser before its API keys became usable. The first API-key call now sets the organization up exactly like a first login would. If your workflow provisions organizations programmatically, the key you mint works immediately.

Clearer recovery when setup fails midway

cleotic setup creates a project, a brand, and a monitor in one go. If a step fails, say the network drops after the project is created and before brands have been created then the CLI doesn't leave you hanging. It tells you exactly what was created, the command to resume where it stopped, and the command to clean up and start over. We deliberately do not auto-delete anything on failure. The resources created remain valid and the process can be resumed.

Behind the scenes: the whole journey, tested on every change

We built a customer-journey test harness that runs on every code change to the CLI or the API. It compiles the CLI binary and walks the full lifecycle a customer would: create a project, set it as default, add a brand, configure a monitor, write a prompt, run it, export benchmark evidence, exercise the delete safeguards, and clean everything up.

None of this changes what you see in the product today. But it improves how confidently we can ship the next thing.

That's all for this week. If you're scripting against the CLI and hit anything rough then send it my way this surface is built for you. Have a great weekend, and we'll see you next time.

See how your brand appears in AI answers

Run a free visibility audit and learn how AI assistants describe your company, competitors, and category.