Home › Running AI agents on flat-rate AI subscriptions
Running AI agents on flat-rate AI subscriptions
Affiliate disclosure: Mulkern AI Systems is an independent affiliate partner of Easy (easy.one), a product of Smart Apps Limited. If you buy Easy through our links, Mike Mulkern earns a commission, and the coupon applied at checkout (MICHAEL) is his. Everything on this page describes how we run our own businesses on Easy. It is our experience, not a statement by Easy.
Our Easy installation runs on two flat-rate AI subscriptions we already pay for: a Claude plan that does the work and a smaller ChatGPT plan held in reserve. Neither bills per token, so the limit we manage is the weekly allowance. This page explains the daily token target, the hourly check, and the habit that mattered more than which model we picked.
Running our Easy installation on a Claude plan is Mike's own arrangement, not a promise that you can do the same. In April 2026 Anthropic changed how Claude subscriptions can be used with third-party tools, so check Anthropic's current terms, or plan on an API key, before you rely on one.
Bring your own AI subscription
Easy runs on the AI subscription or API key you already hold, and Easy adds no per-token charge. You can connect more than one provider, and if one is rate limited or down, the next takes over in the order you set.
We keep the reserve plan for when the main one is in cooldown and a project has to finish. Using the reserve for routine work would leave nothing in hand when it is needed.
A daily target and an hourly check
Mike set a target of roughly one seventh of the weekly allowance per day. A scheduled script checks usage every hour, and a second check runs every two hours. Both are steering signals, not hard stops.
Escalation is based on week-to-date usage, never a single day or a projection. One heavy day is normal. A heavy week is worth a conversation.
Session length matters more than model choice
Most of our daily token use turned out to be conversation history re-sent on every turn of long chats, not the new work itself. The Easy docs on tokens and caching explain why a long chat costs more with every message.
So the rule is shorter sessions on the standard model, not cheaper models. A weaker model produces work that has to be redone. We start a new chat with a short brief for each new topic, save what matters to memory, and turn anything repeated into a skill.
Limit how many agents run at once
Each worker on a job carries its own full context, so width multiplies cost. After one large build with six or more parallel workers used a large share of a day's allowance in three hours, Mike set a cap: no job runs more than four workers at once without his explicit yes.
Sequential stages with tight briefs cost less than wide fan-out, and they are easier to check.
In the Easy documentation, see Tokens and caching and Connect providers.
Questions
Does Easy charge per token?
Easy states that it adds no per-token charge. You pay your AI provider directly through your own subscription or API key.
Which AI providers can Easy use?
Our main page lists Claude, ChatGPT, Kimi, and local Ollama. Check Easy (easy.one) for the current list.
What happens when a provider hits its limit?
With more than one provider connected, Easy can move to the next one in the order you set. We keep a second subscription for exactly that case.
Is this page sponsored by Easy?
No. It is an affiliate page written by Mulkern AI Systems. Mike earns a commission on purchases made through our links.
More from how we run on Easy
Easy is the platform both of our companies run on. Our main page explains what it is, compares the three plans, and applies our coupon at checkout.
See the plans on our Easy page