Aetherfy billing and spend caps
The Aetherfy plans
| Plan | Price | Agents | Max memory | Regions | Always-on | Log retention |
|---|---|---|---|---|---|---|
| Free | $0/mo | 1 | 256 MB | 1 per workspace, chosen when you create it | Runs until 75% of the month is used, then paused — not billed | 3 days, up to 50 MB |
| Starter | $35/mo | Unlimited | 2 GB | 1 per workspace, chosen when you create it | Available up to 2 GB — counts as usage | 14 days, up to 250 MB |
| Performance | $149/mo | Unlimited | 4 GB | Up to 3 per workspace — extra regions count as usage | Available up to 4 GB — counts as usage | 30 days, up to 1 GB |
| Enterprise | Custom | Unlimited | 8 GB | All + custom | Included | Negotiated |
The log column is the window and the most log bytes an account keeps; whichever is reached first applies. The size counts each line’s message bytes, at least 64 for a short line. When an account is over its size the oldest lines are removed first, and because the cleanup runs hourly an account may run over its size for up to an hour. See Limits.
Multi-region placement in Aetherfy begins at the Performance plan. On Free and Starter a workspace has a single region, chosen when it is created — see Regions and replication.
Current pricing detail is published at https://aetherfy.com/pricing .
What Aetherfy charges beyond the plan price
| Plan | Overage |
|---|---|
| Free | No overage — the agent is paused once it is awake beyond 75% of the month, and it resumes within a day of the new month starting (UTC) |
| Starter | Extra agents from $3.50/mo each, by size; always-on add-on $7–$23.10/mo per agent, depending on memory; usage limit approximately $122/mo |
| Performance | Extra agent-regions from $3.50/mo each, by size; always-on add-on $7–$49/mo per agent-region, depending on memory; overage flows up to a usage limit of approximately $443/mo, with warnings as you approach it |
| Enterprise | Negotiated per contract |
The Free plan does not overflow into charges. When a Free account reaches an Aetherfy limit, the action is refused rather than billed.
A Free account has a monthly allowance of awake time. The one auto-sleep agent Free allows is priced like any other, and the allowance is that price; it is metered over a calendar month (UTC), not over a Stripe period, because a Free account has none. Free has no overage: once the agent is awake beyond 75% of the month (see Uptime and the price of an auto-sleep Aetherfy agent), Aetherfy pauses it and emails you. Nothing is billed for it, now or later. The agent resumes within a day of the new month starting (UTC): the control plane’s daily run lifts the pause, and its usage starts again from zero. An agent that sleeps when it is idle stays inside the limit.
Free allows always-on (keep_alive: true) under the same rule, and bills nothing for it. An
always-on agent stays awake continuously, so it reaches 75% of the month after about three weeks,
is paused, and resumes within a day of the new month starting. The pause is not an instant stop at
the line: Aetherfy books awake time once a day, after the day has closed, so the agent can run past
75% by about a day before it is paused. The email Aetherfy sends when it pauses the agent says the same.
While the agent is paused, new deployments are paused with it, as on a paid plan at its spend limit (a code-only change is refused too). Both resume within a day of the new month starting (UTC), or at once if you upgrade: moving to a paid plan lifts the pause and the block on deployments as soon as the upgrade completes, with nothing to do afterwards. A deployment refused while the agent is paused says so in plain words: the agent was awake beyond 75% of the month, so new deploys are paused too, and both resume with the month or now if you upgrade.
If an account on a paid plan with an always-on agent moves to Free, the agent stays always-on and follows this rule from then on.
What each Aetherfy plan includes
The plan price covers a fixed amount of usage each month, counted at the prices in the table below. “Unlimited” agents on the plans above means Aetherfy sets no cap on how many you run; what you run is priced, and the part of it beyond the included amount is the overage.
| Plan | Included each month |
|---|---|
| Free | $3.50 of agent usage — the one 256 MB auto-sleep agent Free allows, awake up to 75% of the month; nothing is billed beyond it |
| Starter | $46.20 of agent usage |
| Performance | $210 of agent usage |
| Enterprise | Negotiated per contract |
A month’s bill is the plan price plus the usage beyond that amount, up to the plan’s usage limit. On Starter, $60 of usage in a month is billed as $35 + ($60 − $46.20) = $48.80.
What an extra agent costs
Usage beyond a plan’s included amount is billed at one rate across every plan, so what one extra agent (per region) costs depends only on its memory size and on whether it is always-on. Starter offers sizes up to 2 GB, Performance up to 4 GB. Per month:
| Memory | Auto-sleep | Always-on |
|---|---|---|
| 256 MB | $3.50/mo | $10.50/mo |
| 512 MB | $4.90/mo | $14/mo |
| 1 GB | $7/mo | $21/mo |
| 2 GB | $15.40/mo | $38.50/mo |
| 4 GB | $35/mo | $84/mo |
What counts as usage in Aetherfy
Four things accrue as usage on an Aetherfy account:
| Source | Notes |
|---|---|
| Agents beyond the plan’s included amount | Per agent, per month |
| Agent-regions beyond the plan | An agent in 3 regions consumes 3× |
| Always-on | Priced by the agent’s memory on a paid plan; not billed on Free, where the agent is paused at 75% of the month |
| Task agents | Priced by memory, per region, exactly like any other agent |
The agent-region rule is the one that catches people out: on Aetherfy the unit is the agent-region pair, not the agent.
How Aetherfy prices a task agent
A task agent is priced the same way every other Aetherfy agent is: by its memory, per region, per month. There is no separate rate for task compute and no separate allowance to keep track of — one plan, one usage figure.
Its machine sleeps between runs and wakes for one, so the time it spends running is treated exactly like the awake time of an agent that is not always-on: see Uptime and the price of an auto-sleep Aetherfy agent below for how that works and when it changes what you pay. An agent that runs tasks occasionally costs what its size costs; one whose tasks keep its machine awake continuously converges on the always-on price for that size, which is the same rule a service agent runs under.
Spawned task agents follow it too, priced by their own memory.
Your live usage for the current period, per agent, is on the usage page at https://app.aetherfy.com/dashboard/billing .
When Aetherfy charges you
The Aetherfy billing period has a deliberate offset between its two components:
| Component | When it is charged |
|---|---|
| Flat plan fee | At the start of the period |
| Usage | At the end of the period, collected on the next period’s invoice |
So a statement shows the base fee for the current period plus usage from the previous one. This one-month offset is intentional, not a lag or a billing error.
Your live usage for the current period is visible at https://app.aetherfy.com/dashboard/billing .
When an Aetherfy billing period starts and ends
Your Aetherfy billing period runs on your subscription’s anniversary, not the calendar month. A subscription started on the 14th bills on the 14th, and its period is the span between those dates — so “this month’s usage” means the current anniversary period, which is usually not the current calendar month. A Free account has no subscription, so its period is the calendar month (UTC).
All Aetherfy timestamps are UTC. If you are reconciling usage against your own records, line them up on the period boundaries shown on the usage meter rather than on month ends.
Unused Aetherfy allowance does not roll over
Whatever your plan includes is per period. Anything you do not use expires when the period closes and does not carry into the next one, so a quiet month does not buy headroom for a busy one.
Changing configuration part-way through an Aetherfy period
Usage is accrued daily, not decided once at the start of the period. Anything that changes what an agent costs re-prices it from that day forward, and the period’s total is the sum of those days:
| Change | Effect |
|---|---|
| Enabling or disabling always-on | Re-prices from that day |
| Adding or removing a region | Re-prices from that day |
| Changing an agent’s memory | Re-prices from that day |
| Deploying or deleting an agent | Counts for the days it existed |
There is no proration penalty and no minimum term on any of these — turning something off part-way through the period stops it accruing from then on. That is why shrinking your way back under a spend cap works immediately rather than at the next cycle.
Uptime and the price of an auto-sleep Aetherfy agent
Awake time. An auto-sleep agent’s price covers it being awake up to 75% of the month, in each region it runs in. Awake time beyond that counts toward your plan’s included usage, like any other usage, at a rate that brings an agent awake all month to the always-on price for its size. You pay extra only once your plan’s included usage for the month is used up. On the Free plan, an agent is paused instead once it is awake beyond 75% of the month, and it resumes within a day of the new month starting (UTC).
An agent that is not always-on is priced on the assumption that it sleeps some of the time, and Aetherfy holds that assumption as a monthly budget rather than a daily one. While an agent’s awake time over the period stays inside that budget, it costs what the price card says.
The budget is a share of the time the agent-region exists over the whole billing period: from the start of the period (or from when the agent was created or the region added, if that was later) to the end of the period. Nothing extra accrues before that budget is used up, however the awake time is spread over the month. An agent-region that exists for only part of the period has a budget for only that part. When one stops existing early (the agent is deleted or archived, or the region is removed), its budget shrinks to the time it existed and the awake time beyond it is settled once, that day.
An agent that stays awake beyond the budget accrues gradually toward the always-on price for its size instead. This is a smooth, proportional effect, not a cliff and not an enforcement action: on a paid plan nothing is throttled, paused or suspended for being awake. Free is the exception: its agent is paused once it is awake beyond 75% of the month, until it resumes within a day of the new month starting (UTC) (see What Aetherfy charges beyond the plan price). An agent that is genuinely awake all the time ends up costing what always-on costs, which is the point — the two ways of being continuously available are priced the same.
Because the budget is monthly, individual busy days do not matter. An agent that spikes to full uptime on a few days and idles the rest stays inside its budget and its price does not change.
A fan-out spends that budget faster, and this is the one case where the arithmetic surprises people. Spawning the same task agent several times over runs each spawn on its own machine, and awake time is summed across those machines, not averaged over them: three machines awake for a minute is three machine-minutes of budget, the same as one machine awake for three. The monthly price card charge for the agent-region does not change and the extra machines are released once they go idle, so a fan-out costs you awake time and nothing else. How wide a fan-out can go at once is the runs-in-flight row on Limits.
A redeploy does the same thing in time rather than in width. The replaced version
stays up for about a minute after the new one serves, and longer while a call is open.
It is billed as awake time until the machine stops, on top of the new version’s machine,
and it stops once the call has ended and a quiet minute has passed, or at the
drain_timeout_minutes maximum at the latest (/agents).
Which size the time beyond the budget is priced at
When you deploy a different memory size during a period, the time beyond the budget is priced at the size of the release that pushed it there, never at whichever release happens to be newest when the day is booked. Precisely: the excess a day adds is priced, level by level, at the release that was awake when the excess last rose to that level; when two releases run at once — the old one and the new one during a rollout — they share it in proportion to how many machines each has awake. Excess that rises and falls back within the day is not billed twice; what remains at the day’s end is priced where it was last reached. So a release that is quiet in the morning and replaced by a busy one at night has the day’s excess priced at the new size, because the night is what took it past the budget. A deploy cannot re-price time that already ran at another size, in either direction.
A run is priced at the size of the release it runs. A release whose rollout failed is priced, for the time its machines were awake, at the size it was built with; it adds no monthly price card charge, because it never went live.
The corollary is worth stating plainly: always-on is a runtime promise, not a billing exemption. Turning it on asks Aetherfy to keep the agent warm; it does not change how the agent is metered. An always-on agent that really does stay awake bills its always-on rate either way. On Free, where nothing is billed, it is paused once it has been awake for 75% of the month.
Changing your Aetherfy plan
Upgrades and downgrades behave differently on purpose.
| Direction | When it takes effect | Charge |
|---|---|---|
| Upgrade | Immediately | Prorated for the remainder of the current period |
| Downgrade | At the end of the current period | No charge, and no refund |
An upgrade applies at once — you get the higher limits straight away and pay the difference for the days remaining. A downgrade is scheduled rather than applied: you keep the plan you paid for until the period ends, then move down. Nothing is refunded, because nothing is cut short.
A scheduled downgrade can be cancelled before it lands, at https://app.aetherfy.com/dashboard/settings/billing , which leaves you on your current plan.
Aetherfy only downgrades an account whose resources fit the target plan. That is checked twice: when you schedule the downgrade, and again shortly before the renewal that would apply it — because in between you are still on the higher plan and can legitimately grow past what the lower one allows. If the account no longer fits at the second check, the downgrade does not land: your current plan continues, and Aetherfy emails you the specific resources that are over the target plan’s limits. Bring them within the limits and schedule the downgrade again. If Aetherfy cannot complete that second check, the scheduled downgrade is left as it is and re-checked on the next pass, so it is neither cancelled nor accompanied by an email.
Agent count, memory, regions and workspaces are all lower on lower plans, and the dashboard tells you what would be over the line before you schedule anything.
Cancelling an Aetherfy subscription
Cancelling sets the Aetherfy subscription to end at the current period’s end. The subscription runs to that date with full service, and you can resume before it without losing anything. Usage accrued during the final period is invoiced at period end, following the same offset described above.
Failed payments on an Aetherfy account
After a failed payment, Aetherfy applies a 14-day grace period. During it,
existing agents keep running while the payment issue is unresolved, and Aetherfy
freezes growth with HTTP 403 and the code DUNNING_FROZEN: new spawns and task
runs are refused, and a deploy goes through only if it reduces cost (see
Deploying while growth is frozen). Growth
resumes once a payment succeeds, and the freeze clears automatically — there is
nothing to re-enable by hand.
If the grace period ends without a successful payment, Aetherfy suspends the account, and that is the point at which running agents stop. Suspension stops machines; it does not delete anything:
| What happens | Detail |
|---|---|
| Your agents | Machines are stopped, they no longer wake on incoming requests, and each agent’s status becomes stopped — see Runs and logs |
| API calls | Refused with HTTP 403 ACCOUNT_SUSPENDED |
| Your data | Untouched. Collections, secrets and stored code all survive |
Settling the account reverses it: Aetherfy restores how those agents’ machines
wake on requests and starts a service’s machines again. A service returns to
running once one of its machines has started; a task agent returns at once, and
its machines start with its next run. Nothing needs recreating, and no
configuration is lost.
Update the payment method at https://app.aetherfy.com/dashboard/settings/billing .
Setting your own spend cap in Aetherfy
You set your own limit at https://app.aetherfy.com/dashboard/settings/billing .
The Aetherfy spend cap is expressed in total-bill dollars — base price plus usage, not usage alone.
| Rule | Value |
|---|---|
| Minimum | Strictly above your plan’s base price |
| Maximum | At or below your plan’s system limit |
| System limit, Starter | approximately $122 |
| System limit, Performance | approximately $443 |
| Free and Enterprise | No spend-cap control |
A cap at or below the base price is rejected by Aetherfy, because the running total starts at the base price on day one of every period; such a cap would pause the account instantly.
What Aetherfy does when the limit is reached
There are two distinct stages, and the difference matters.
Stage 1 — growth freezes. Aetherfy blocks new spawns and new task runs, and
a deploy goes through only if it reduces cost (see
Deploying while growth is frozen). Your
existing agents keep running. You can always shrink your way back out. The API
returns HTTP 403 with the code SOFT_CAP_EXCEEDED.
Deploying while growth is frozen
Both freezes, the spend limit (SOFT_CAP_EXCEEDED) and a failed payment
(DUNNING_FROZEN), apply one rule to every deploy: an upload, a git push, a
rollback, a redeploy and a move. The deploy goes through only if it reduces
cost and grows nothing, measured against what is live now. aetherfy.yaml
sets how an agent runs, so a deploy is how you lower it.
Allowed, alone or together, as long as nothing grows:
- turning always-on off (
keep_alive: false); - lowering
memory_mb; - deploying to fewer regions;
- shortening
idle_timeout_minuteson a service that auto-sleeps.
Refused:
- anything that grows: always-on on, more memory, more regions, a longer idle timeout;
- a deploy that changes none of the four: the same configuration, a code-only change, a rollback or redeploy that reduces nothing, a move to another workspace, or the same number of regions in different places;
- a rollback or redeploy to a release whose configuration grows anything — a release runs the configuration it was deployed with, so rolling back to a bigger one is growth, and rolling back to a smaller one is a reduction;
- a mix of a reduction and a growth;
- an agent’s first release, or a deploy of an agent with nothing live.
The refusal names any setting that would grow and ends with: “While your account is frozen a deploy must reduce cost and grow nothing: turn always-on off, lower memory, use fewer regions or shorten the idle timeout.”
Stage 2 — machines pause. Aetherfy pauses machines only if you are also past your plan’s included usage. Within the included allowance the runtime is already paid for, so pausing would save nothing while denying you service you have already bought.
Raising or removing the cap resumes Aetherfy service immediately. It does not wait for the next billing cycle.
Scheduled tasks while an Aetherfy account is over its limit
While an Aetherfy account is over its limit, scheduled task occurrences are recorded as missed rather than firing. Nothing needs re-enabling: the next occurrence fires normally once you are back under the limit. Missed occurrences are not backfilled, so a task that should have run five times during the freeze runs once, at its next scheduled time.
Deploy-time cost confirmation in Aetherfy
At deploy time, an Aetherfy deployment that would add usage prompts for
confirmation first. The API answers HTTP 402 with the code
OVERAGE_CONFIRM_REQUIRED, naming the added monthly amount.
This is informational, not a block. The CLI’s --yes flag proceeds past it — see
afy deploy.
Aetherfy has no deploy rate limit
There is no per-hour deploy throttle in Aetherfy. Deploying repeatedly is not itself restricted; the spend cap is the only bound on deploy-driven cost.
Where to find billing information in the Aetherfy dashboard
| What | Where |
|---|---|
| Usage and spend meter | https://app.aetherfy.com/dashboard/billing |
| Plan, payment method, spend cap | https://app.aetherfy.com/dashboard/settings/billing |
| Public pricing | https://aetherfy.com/pricing |
| What each billing screen shows | Usage and billing in the dashboard |
Aetherfy billing FAQ
When am I charged?
Aetherfy charges the flat plan fee at the start of each billing period and charges usage at the end of the period, collecting it on the following period’s invoice. That means a single statement mixes two periods: the base fee for the period beginning now, and the usage accrued during the period that just closed. The one-month offset on usage is intentional. If your first invoice looks like it is missing usage, it is not — that usage will appear on the next one.
What happens when I hit my spend limit?
Two things happen in sequence, and only the first is guaranteed. First, growth
freezes: Aetherfy refuses new spawns and new task runs, and any deploy that does
not reduce cost, returning HTTP 403 SOFT_CAP_EXCEEDED. A refused deploy names
the first setting that would grow in detail.field. Your existing agents keep
running throughout, and a deploy that only reduces cost is still permitted (see
Deploying while growth is frozen). Second, machines pause — but only if you are also past your
plan’s included usage. Inside the included allowance, Aetherfy leaves your
machines running, because that runtime is already paid for and pausing it would
save you nothing. Raising or removing the cap resumes service immediately rather
than at the next cycle.
Can I set my own spend limit, and what range is allowed?
Yes, on Starter and Performance, at https://app.aetherfy.com/dashboard/settings/billing . The Aetherfy cap is expressed in total-bill dollars — base price plus usage — and must be strictly above your plan’s base price and at or below your plan’s system limit, which is approximately $122 on Starter and approximately $443 on Performance. Aetherfy rejects a cap at or below the base price because the running total starts at the base price every period, so the account would pause the moment the period opened. Free and Enterprise accounts have no spend-cap control.
Is there a deploy rate limit?
No. Aetherfy enforces no per-hour deploy throttle, and there is no cooldown
between deploys. The only bound on deploy-driven cost is your spend cap, which is
why setting one is worthwhile if automation in your pipeline can trigger deploys
without a human in the loop. Note that a deployment which would add usage still
prompts for confirmation with HTTP 402 OVERAGE_CONFIRM_REQUIRED — that is a
cost disclosure, not a rate limit.
My payment failed — what happens?
Aetherfy gives you a 14-day grace period. During it your existing agents keep
running, but growth is frozen: new spawns, task runs and any deploy that does
not reduce cost are refused with HTTP 403 DUNNING_FROZEN (see
Deploying while growth is frozen). Fix the payment method at
https://app.aetherfy.com/dashboard/settings/billing ; once a payment succeeds
the freeze clears automatically and deploys resume. You
do not need to contact anyone or re-enable anything by hand. If the 14 days
elapse without a successful payment the account is suspended, which stops your
agents’ machines and moves each agent to the stopped status — reversed, with
no data loss and nothing to recreate, as soon as the account is settled.
Suspension is not indefinite. See the next answer for what happens if it is never settled.
How long does a suspended account keep its data?
Thirty days. A suspended account keeps everything for 30 days after the suspension begins — day 44 of the lifecycle, counting from the failed payment — and settling the balance at any point in that window restores the account completely, with nothing to recreate.
If the 30 days elapse without payment, your running agents and your vector collections are permanently deleted. This is the one irreversible step in the lifecycle: the vectors are gone, and so is anything the deleted agents were holding. Aetherfy cannot restore them afterwards, on request or otherwise.
Your account itself is not deleted. Your account and your login remain, so you can sign back in and resubscribe at any time. Everything that ran or stored data is removed: your agents and their machines, their deployed apps and code bundles, and your vector collections. None of it comes back when you resubscribe — you recreate your agents and collections from scratch.
Your credentials are removed with them. The secrets you stored for your agents and workspaces are deleted, and your API keys are revoked: requests made with them are refused. Aetherfy also removes its record of your GitHub connection and stops using it, but the Aetherfy GitHub App stays installed on your GitHub account until you remove it there. To revoke its access completely, uninstall the App from your GitHub settings.
You are warned before it happens. Seven days ahead of the deletion Aetherfy emails you a final notice saying how many days remain, and paying at any time before the deletion runs stops it. Before the deletion you receive five emails in all — the failed payment, two reminders during the grace period, the suspension, and the final notice — and a sixth confirms the deletion once it has happened.
Does always-on cost extra?
Yes, on every paid plan where it is available. On Free it costs nothing: the agent stays awake until it has been awake for 75% of the month, then it is paused until the month ends (see What Aetherfy charges beyond the plan price). On a paid plan the add-on is what always-on costs on top of the same agent on auto-sleep, so it depends on the agent’s memory. On Starter it is $7–$23.10/mo per agent, and always-on is available up to 2 GB. On Performance it is $7–$49/mo per agent-region, and always-on is available up to 4 GB. On Enterprise it is included. In all cases where it is an add-on, always-on counts as usage in Aetherfy, which means it accrues against your spend cap and against your plan’s included usage.
Does a 3-region agent cost 3×?
Yes. The billable unit in Aetherfy is the agent-region pair, not the agent, so an agent deployed in three regions consumes three of them. On Performance, agent- regions beyond the plan’s inclusion are billed by size, from $3.50/mo each (see What an extra agent costs). This is the single most common surprise on an Aetherfy invoice, and it is worth checking your region list before scaling agent count.
I removed a region from aetherfy.yaml — why am I still billed for it?
Because removing a region from aetherfy.yaml is a compute-only change in
Aetherfy. The machines in that region come down, but your vector data remains
there and the replica keeps being billed until you explicitly release it. Editing
your deployment configuration never deletes data — that is deliberate, since a
config edit silently destroying a region’s vectors would be far worse. To stop
being billed for the data, release the replica explicitly. See
Regions and replication.
What happens if I cancel mid-period?
Cancelling sets the Aetherfy subscription to end at the current period’s end rather than terminating it on the spot. Your agents keep running to that date, and you can resume before it without any loss of state. Usage accrued during the final period is invoiced at period end, following the same one-month offset as any other period, so expect one final statement after the subscription has ended.