Plans and billing
Free costs nothing and has daily caps. Pro is $49 a month with a monthly allowance of runs, steps, CPU time, outbound requests and model use, then overage up to a spending limit you set.
Free and Pro
Every workspace starts on Free. An admin upgrades to Pro in Settings → Plan and billing.
| Limit | Free | Pro |
|---|---|---|
| Price | $0 | $49 a month |
| Runs | 1,000 per 24 hours | 250,000 a month, then $0.10 per 1,000 |
| Outbound requests | 3,000 per 24 hours | 1,000,000 a month, then $0.02 per 1,000 |
| Steps | 5,000 per 24 hours | 1,000,000 a month, then $0.02 per 1,000 |
| CPU time | 300 seconds per 24 hours | 50,000 seconds a month, then $0.05 per 1,000 seconds |
| CPU time per run | 10 seconds | 60 seconds |
| Runs at once | 10 | 100 |
| Runs waiting per workflow, past its concurrency | 1,000 | 10,000 |
| Workflows · active triggers | 10 · 20 | 50 · 100 |
| Secrets · connections | 25 · 10 | 250 · 100 |
| TCP targets | not included | 20 |
| Agents the helper isn't counted | 3 | 25 |
| Model allowance agents and the helper | 50,000 tokens per 24 hours, $2 of model use a month | 5,000,000 tokens per 24 hours, $2 of model use a month included, then list price × 1.20 |
| Models | the small models | every model, and models on your own provider key |
| Member invitations per 24 hours | 20 | 200 |
| Seats members and pending invitations | 2: one builder and one approver | 1 builder and 1 approver, then up to 2 more builders at $15 a month each |
| Alert emails per day | 50 | 500 |
| Runner tokens | 2 | 20 |
| Longest command on a runner · commands waiting for a runner | 30 minutes · 20 | 12 hours · 500 |
Larger teams: Team has up to 10 seats of any role, and Business and Enterprise have no seat limit. None of them are in Settings yet; email sales@shiftlabs.org to ask about them.
- Both plans deploy up to 100 new versions per 24 hours, and share the run limits under Limits. Pricing lists every limit, and
GET /v1/mereturns your workspace's. - On Pro, runs, steps, CPU time and outbound requests per 24 hours stop at the month's included amounts: 250,000, 1,000,000, 50,000 seconds and 1,000,000.
- Model use on Free has no overage: past the $2 a month, agent messages are refused until the month ends. On Pro, model use past the month's $2 is overage. See Model use.
What counts
On Pro, a run counts toward the month's runs once it ends, if your code ran.
- Counts
A run that completed, or that failed after your code ran: an error it threw, a step out of retries, or a limit it reached. A run that was cancelled or terminated after its code ran. A re-run is a run of its own, and so is a run an agent's tool call starts.
- Doesn't count
Runs refused at a limit, runs that never started, and runs that failed on our side, were stopped by a suspension or ended with a deleted workspace. Nor does a run cancelled before its code ran, such as a queued or held one, or a held run dropped when its workflow resumes.
- Outbound requests
Every request your workflows send counts, including requests the gateway refuses.
- Steps and CPU time
Each attempt of a
step.runyour code runs counts once, so a retry is one more step: astep.runthat fails twice and then passes counts 3. In a run started before 3 November 2026, astep.runcounts once however often it is retried. Sleeps, and the waits between attempts, don't count. CPU time is the time your code runs on a CPU. Both count in the hour they're used, whether the run then completes or fails. Code stopped at its CPU limit counts the CPU time it used before it stopped, about twice the limit. About 67 stopped calls use Free's 300 seconds a day, and stopped invocations count toward a run's CPU time per run the same way. - Agent messages
Not runs. Their model use counts toward the model allowance, and on Pro toward the month's model use. Their code tools' CPU time counts toward CPU time.
On Free, a run counts toward runs per 24 hours from when it is queued, however it ends. Held and waiting runs count too. Steps and CPU time count toward their caps as they are used, and runs already started continue past them. A run that hasn't started yet, waiting for its workflow's concurrency or its turn in a re-run, or held until a resume, waits while a cap is reached and starts once enough of the 24 hours has passed. GET /v1/usage shows runsLast24h, stepsLast24h and cpuSecondsLast24h, and on Pro, GET /v1/billing shows the month's usage.
Model use
Agents and the helper use models at the prices under Models. Every plan includes $2 of that each month. On Pro, model use past it is overage at the list price × 1.20.
- Each response counts at the list price of the model that answered, to the nearest $0.000001. $1 of list price past the $2 is $1.20 of overage.
GET /v1/billingshows the month's model use asusage.modelMicros, in millionths of a dollar, andincluded.modelMicrosis 2,000,000, the $2.- Model use goes on past the $2 until the spending limit, which counts it with runs, steps, CPU time and outbound requests. While a payment is failing, it stops at the $2.
- Model use stops at $1,000 a UTC month and at 5,000,000 tokens per 24 hours. See Agents.
- On Free, and on Pro until your subscription includes model credit, model use stops at $2 a UTC month; on such a Pro subscription
usage.modelMicrosisnull. An agent message past it returns429 {"error": "quota exceeded", "limit": "modelCentsPerMonth", "max": 200}. From a member or an API key it also carriesdetail, naming the $2 and the reset, andresetsAt, the next 00:00 UTC on the 1st in milliseconds. A chat token gets the short form. - A model on your own provider key is billed by your provider, and adds nothing to model use or overage here. Its tokens count toward the 5,000,000 per 24 hours.
Overage and the spending limit
Past the month's included runs, steps, CPU time, outbound requests and model use, Pro bills overage: $0.10 per 1,000 runs, $0.02 per 1,000 steps, $0.05 per 1,000 CPU-seconds, $0.02 per 1,000 outbound requests, and model use at list price × 1.20. The spending limit caps it. Overage is charged as it builds up: each time another $50 of it accrues in a month, it is invoiced and charged to your card, and the rest is on the monthly invoice. The included amounts aren't given again.
# $250 a month (admin key); 409 above $50 until the subscription's first monthly renewal.
# Answers {"spendingLimitCents", "capped", "cappedBy", "overageCents"}.
curl -X PATCH https://api.shiftlabs.org/v1/billing \
-H "authorization: Bearer $SHIFTLABS_API_KEY" \
-H "content-type: application/json" \
-d '{"spendingLimitCents": 25000}'
- The limit is $100 a month until an admin changes it, in Settings → Plan and billing or with
PATCH /v1/billing, from $0 to $10,000. At $0, Pro stops at the included runs, steps, CPU time, outbound requests and model use. - Until your subscription's first monthly renewal, the limit is at most $50. Above that,
PATCH /v1/billingreturns409and Settings won't save it, and a higher limit saved before, the $100 default included, applies from the renewal.GET /v1/billingshows the most you can set now asspendingLimitMaxCents. - The billing contact, the email given at checkout, gets an email at 80% of the limit and another at 100%.
- At the limit, new runs and agent messages return
429 {"error": "spending limit reached", "detail"}withRetry-After: 3600, webhook deliveries included. Runs already started finish. Runs that haven't started yet, waiting for their workflow's concurrency or their turn in a re-run, or held until a resume, wait. An agent's workflow tool calls fail. - Usage is checked every hour. A workspace can pass its limit by up to an hour of use.
- Raising the limit, or a new billing month, starts new runs again at once, and waiting runs within 5 minutes.
When a payment fails
A workspace whose payment failed stays on Pro for 7 days. Without a payment by then, it moves to Free.
- While a payment is failing there is no overage. New runs and agent messages stop once the month's included runs, steps, CPU time, outbound requests or model use are used, with
429 {"error": "payment failed", "detail"}, and runs that haven't started yet wait. When the payment that failed is a $50 overage charge, they're used already, so new runs stop at once. Raising the spending limit doesn't lift this. - Our payment processor retries a failed payment and emails the billing contact when it fails. When new runs stop, the billing contact also gets one email:
New runs paused for <workspace>: payment failed. - An admin updates the payment method from Settings → Plan and billing. New runs start again once the payment goes through, and waiting runs within 5 minutes.
GET /v1/billingshows"status": "past_due",graceUntiland, once runs stop,"cappedBy": "payment".
Upgrade and manage billing
An admin upgrades from Settings → Plan and billing, which opens a checkout page. Pro starts when the payment goes through.
- When billing is unavailable,
GET /v1/billingreturns"available": false, and checkout and the billing portal return503 {"error": "billing is not available yet"}. - On Pro, Manage billing opens the billing portal: invoices, the payment method, the tax ID and cancelling. Cancelling keeps Pro until the end of the billing month.
- Everyone in the workspace can read the plan, usage and spending limit. Admins set the spending limit and the PO number, see the tax ID and manage billing.
- See Refunds for refunds.
PO number and tax ID
Invoices show your PO number and your business tax ID.
# Upgrade with a PO number (admin key). Answers {"url"} of the checkout page.
curl -X POST https://api.shiftlabs.org/v1/billing/checkout \
-H "authorization: Bearer $SHIFTLABS_API_KEY" \
-H "content-type: application/json" \
-d '{"poNumber": "PO-4471"}'
# Change it later. Answers {"poNumber"}. "poNumber": null removes it.
curl -X PATCH https://api.shiftlabs.org/v1/billing/invoice \
-H "authorization: Bearer $SHIFTLABS_API_KEY" \
-H "content-type: application/json" \
-d '{"poNumber": "PO-4472"}'
- Enter the PO number when you upgrade in Settings → Plan and billing, or send it with
POST /v1/billing/checkout. Every invoice carries it, the first one included. - Admins change or remove it in Settings → Plan and billing on Pro, or with
PATCH /v1/billing/invoiceon either plan, up to 140 characters. Invoices already issued keep the PO number they show. - Outside the US, checkout asks for your business tax ID, and for now Pro needs one before you pay wherever checkout can take it. Add or change it later in the billing portal.
API
Read billing and set the spending limit and the PO number with the /v1 API.
| Route | Method | Minimum role |
|---|---|---|
/v1/billing {available, plan, paid, status, periodStart, periodEnd, cancelAtPeriodEnd, graceUntil, hasBillingAccount, spendingLimitCents, spendingLimitMaxCents, capped, cappedBy, usage}; paid: false on Free; usage on Pro, with modelMicros null until the subscription includes model credit; spendingLimitMaxCents: the most the spending limit can be set to now | GET | viewer |
/v1/billing {"spendingLimitCents"}, 0 to 1,000,000; 409 above spendingLimitMaxCents, 5,000 until the subscription's first monthly renewal | PATCH | admin |
/v1/billing/checkout optional {"poNumber"}, as on PATCH /v1/billing/invoice; {url} of a checkout page for Pro; 409 when already subscribed, or when the billing account has 4 other invoice fields | POST | admin |
/v1/billing/portal {url} of the billing portal; 409 before there's a billing account | POST | admin |
/v1/billing/invoice {poNumber, taxIds}; PATCH {"poNumber"}, up to 140 characters, or null to remove it | GET · PATCH | admin |