Match the tier to the job

Pricing is the lever people get backwards — they buy the top tier "to be safe" and then use 5% of it. Match the plan to the actual job instead.

The shapes, as of late 2026 (confirm current pricing with each vendor before committing):

• Meta Muse — free for most use, then Power ($20/mo) and Maximum ($100/mo) for heavier usage. If your tasks are occasional and light, the free tier may genuinely be enough; move up only when you hit a wall.
• OpenAI Dots — no free tier; one dot comes with the ~$100/mo ChatGPT Pro plan (and with Business Premium). So Dots is a fit mainly if you're already paying for Pro or would use it anyway.
• Grok Bot — bundled into paid Grok / SuperGrok tiers (reported from roughly $30/mo at the base, more for heavier usage pools). You're choosing a usage pool, not a per-task price.

Two rules keep you from overpaying:

1. Start at the lowest tier your workload plausibly fits, and let a real limit push you up. It's easy to upgrade later; it's wasteful to sit on headroom you never touch. A limit you actually hit is real information; "might need it someday" is not.
2. Count the agent's work honestly before you scale the plan. If you're hitting a tier's limit, first check why — often it's a too-frequent schedule or an over-broad goal (Module 2's lessons), not genuine demand. Fixing the task is cheaper than buying the next tier.

And the availability reminder that overrides price entirely: Muse is US-only and Dots isn't in the EEA/UK/Switzerland yet, so for some people the cheapest usable option is simply the one that works where they are.