Agent plan

The Agent plan is a monthly subscription that pays for your requests instead of your balance, up to a 5-hour and a weekly usage limit. It works with the same API key, the same endpoints and every model; when a limit is reached, your requests continue on your balance if you allow it, or wait for the next reset. The current tiers, their monthly prices and their limits are on Pricing.

If you only call the API now and then, pay as you go is simpler: add funds and pay per token. The plan suits work that runs all day — coding agents, a monitoring loop that sends images every few seconds, a nightly pipeline.

What does the Agent plan cover?

Every request you make with your API keys on all three protocols — OpenAI-compatible (Chat Completions), OpenAI Responses and Anthropic-compatible (Messages) — on every model.
Requests you send from the Playground.
Not the support chat: a paid support conversation is always billed to your balance.

A request that fails is not charged and counts nothing toward your limits.

How do the 5-hour and weekly limits work?

Each tier has two dollar limits, and the plan counts your usage against both:

The 5-hour window starts with your first request on the plan and lasts five hours. When it ends, the window is empty again, and your next request starts a new one.
The weekly window runs in whole weeks from the moment your plan started. At the end of each week it starts again from zero.

The last request before a limit can take you slightly past it; the next request is then handled as described below.

What happens when a limit is reached?

That depends on the setting At a limit, continue on my balance on the Plan page. It is off by default: turn it on to keep working on your balance when a limit is reached.

On: your requests continue on your balance, at the normal per-token prices, until the window resets. If your balance is empty too, you get the usual "add funds" error. You can also set a limit on this balance use for each billing period (see below).
Off: the request is refused with HTTP status 402 and a Retry-After header. Its value is the number of seconds until the earliest reset of the limit you reached, so an agent loop can sleep for that long and try again.

These are example 402 responses (documentation of what the API returns, not a request to run). On the OpenAI-compatible (Chat Completions) and OpenAI Responses endpoints, with Retry-After: 3600:

{  "error": {    "message": "Your plan reached its limit. It resets in 3600 seconds; turn on 'continue on my balance' or add funds to keep going.",    "type": "insufficient_quota",    "code": "plan_limit_reached",    "param": null  }}

On the Anthropic-compatible (Messages) endpoint, with Retry-After: 125:

{  "type": "error",  "error": {    "type": "billing_error",    "message": "Your plan reached its limit. It resets in 125 seconds; turn on 'continue on my balance' or add funds to keep going."  }}

The message names the setting by its current label; the status (402), the type, the code and the Retry-After header are unchanged. The type is the same as for an empty balance, so existing error handling keeps working; the code plan_limit_reached (or the message, on the Anthropic-compatible (Messages) endpoint) tells the two apart. The full error tables are in the API reference.

Can I limit how much of my balance is used beyond the plan?

Yes. On the Plan page, the card Balance use beyond your plan holds the setting At a limit, continue on my balance and a field Limit per billing period (US$). Leave it blank for no limit (the default), or enter an amount and press Save. The field can only be changed while the setting is on. Under it, Used this period shows how much of your balance went to requests beyond the plan in the current billing period, out of your limit. The amount starts again from zero when your plan renews.

Requests beyond the plan are billed to your balance exactly like any other request, and on the Usage page their cost carries a Beyond plan tag. When the amount used this period reaches your limit, a request beyond the plan is refused with HTTP status 402, the code overflow_cap_reached and a Retry-After header. Its value is the number of seconds until the earliest reset of your plan's limits, when the plan pays again. Raise the limit on the Plan page, or wait.

With a limit set, the card warns you under Used this period once this period's use reaches 80% of the limit, and again when the limit is reached. While your plan is at its limit, the account sidebar and the overview's plan card show a Beyond plan line with the same amount, in the warning colour from 80%. Tokener does not send an email for this.

Example responses (documentation, not a request to run). On the OpenAI-compatible (Chat Completions) and OpenAI Responses endpoints, with Retry-After: 3600:

{  "error": {    "message": "Your Agent plan is at its limit and you reached your limit for balance use beyond the plan this billing period. It resets in 3600 seconds; raise the limit on the Agent plan page or wait.",    "type": "insufficient_quota",    "code": "overflow_cap_reached",    "param": null  }}

On the Anthropic-compatible (Messages) endpoint, with Retry-After: 125:

{  "type": "error",  "error": {    "type": "billing_error",    "message": "Your Agent plan is at its limit and you reached your limit for balance use beyond the plan this billing period. It resets in 125 seconds; raise the limit on the Agent plan page or wait."  }}

How is plan usage valued?

At exactly what the request would have cost from your balance: the model's per-token prices, cached input at its cached price where the model lists one, and the data-handling discount when your account has it. So a dollar of plan usage is a dollar of the price list. How the numbers are counted and billed explains the formula.

How do I keep one key on the balance?

On the API keys page, each active key has a Funding menu with two choices:

Plan first (the default) — the key uses your Agent plan while it has room, then your balance.
Balance only — the key never uses the plan and is always billed per token from your balance. Use it for a production app you want billed per token while your agents use the plan.

The change applies to the key's next request.

Where do I see my plan?

The Plan page shows your tier, its status and renewal date, and two bars — 5-hour window and Weekly window — with what you have used of each limit. Under the 5-hour bar it says when that window resets, or "Starts with your next request" when no window is open; under the weekly bar it says when the week resets. A limit you have reached is marked At the limit.
The Usage page lists plan-funded requests like any other, with a Plan tag next to the cost: that cost is the plan's valuation of the request and was not charged to your balance.
Your plan payments (the first charge, renewals and prorated upgrades) are listed under Transactions → Payments, each with its Stripe invoice; requests the plan covers are never charged to your balance; Balance activity lists them per day as Plan usage (not charged), for reference.
In the Playground, each reply's stats line ends with Agent plan when the plan covered it (or Balance when your balance paid), and the sidebar's 5-hour line updates within a moment.

On Pricing, the Agent plan tab lists the tiers; choose one. This takes you to the Plan page in your account (sign in first if you are not signed in). In your account, Agent plan in the sidebar opens the same page.
On the Plan page, press Choose on the tier you want. Payment is handled by Stripe Checkout; the plan renews every month until you cancel.
To cancel, change the card or download invoices, press Manage billing on the Plan page. It opens the Stripe billing portal. A cancelled plan keeps working until the end of the period you paid for. Agent plan payments are non-refundable, except where required by law. If you cancel, your plan stays active until the end of the period you paid for, and you are not charged again.

If a renewal payment fails, the plan keeps working for a short grace period while you update the card, then stops. The Plan page shows the date it stops.

How do I upgrade or downgrade?

On the Plan page, the Change plan section lists the other tiers you can move to, each with its own button:

Upgrade to and the tier's name (a tier with a higher monthly price) starts now. Stripe shows the prorated difference for the rest of the current billing period; you pay it when you confirm, and the new tier's limits apply at once.
Switch to and the tier's name (a tier with a lower monthly price) takes effect at the end of the current billing period, and you keep your current tier until then. The Plan page then says when your plan changes, and Keep with your current tier's name undoes the change.

Your 5-hour and weekly windows keep running through a change: the usage already counted stays, and the new tier's limits apply to the running windows. Only one change can be scheduled at a time, and a cancelled plan has to be resumed in Manage billing before you can change tiers.

Limits

One Agent plan per account at a time.
Plans are priced and charged in US dollars.
A plan belongs to one account and is not shared with other accounts.
The support chat is always billed to your balance, never to the plan.

Pricing — the current tiers and their limits.
API reference — error codes, including plan_limit_reached.
What is Tokener? — balance, keys and protocols in plain words.
FAQ — short answers about pricing and billing.