Skip to content

g1t/apps/docs/src/content/docs/guides/usage-and-billing.md

380 lines17,248 bytesCodeBlame
1---
2title: Usage and billing
3description: What g1t costs, how a workspace pays, the limits that keep unpaid usage in check, and enterprise billing.
4---
5
6Hosting repositories, issues, pull requests, review and your own agent cost
7nothing on g1t. What costs money is what g1t runs for you: its agents'
8models, the sandboxes they and your checks run in, and deployed apps. Each
9is charged at what it costs g1t plus a set markup, after it is used, to the
10workspace that owns the repository. There is no seat price.
11
12Some features are paid for with a monthly plan the workspace turns on, and
13are never free, including while the rest of g1t is. See
14[Plans](#plans).
15
16## Plans
17
18A plan turns on one paid feature for the whole workspace: a monthly price
19that includes an allowance, with usage past it charged at cost plus 20%.
20
21| Plan | Price | Includes each month |
22| --- | --- | --- |
23| [Deployments](/guides/deployments/) | $5 a month | 10 apps up at once, 1 million requests, 3 million CPU milliseconds. Builds are charged by the second. |
24
25Only an owner can turn a plan on or off.
26
271. Open **Settings → Billing**, `g1t.sh/<workspace>/-/billing`.
282. Under **Plans**, choose **Turn on Deployments**, and pay on the card
29 page you are sent to.
30
31Back on Billing, the plan says **On** and when it renews; the card is kept
32and charged each month. **Turn off at the end of the period** ends the plan
33on its renewal date, with nothing more charged after; **Keep Deployments**
34takes that back. If a renewal payment fails, the plan says **Payment
35failed** and the feature stops until it is paid.
36
37Plan usage past the allowance and builds are drawn from the workspace's
38credit, so a workspace with a plan can add credit while agents are free.
39
40## The free allowance
41
42So anyone can try g1t's agents without a key of their own, every workspace
43gets **$1 of model cost on g1t's own models**, free, until **October 22,
442026** (11:59 PM Pacific). That is roughly 10 to 25 agent runs: changes,
45reviews and revisions. The allowance pays for the model part of those
46runs; everything else, such as sandbox time past the free minutes, is
47charged as usual.
48
49- Mission control and **Settings → Integrations** show what is left.
50- When it is used up, agents and workflow runs stop starting, and the
51 pages that start them say so and link to Integrations. Connect your own
52 model provider there and everything carries on at once.
53- The allowance draws on one shared pool. If the pool runs out before
54 October 22, the allowance ends for everyone early.
55- Workspaces on their own provider never use it.
56
57## What is charged
58
59| | Charged |
60| --- | --- |
61| Making a change for an issue | Yes |
62| Revising a change after checks, a review or a person | Yes |
63| A review by a g1t agent | Yes |
64| Catching up with `main` | Yes, when it needed an agent |
65| Planning an [outcome](/guides/outcomes/) | Yes |
66| Acceptance checks | [Sandbox time](#sandbox-time) |
67| The [merge queue](/guides/merge-queue/) | [Sandbox time](#sandbox-time) |
68| [Workflow](/guides/actions/) jobs | [Sandbox time](#sandbox-time) |
69| [Deployments](/guides/deployments/) | The plan, and builds and usage past it. Never free. |
70| Repositories, git, issues, pull requests, the API and MCP | No |
71
72Each run is charged when it finishes: what the model provider charged for
73it, plus 20%. A small change costs a few cents.
74
75Work a workspace routes to [its own model providers](/guides/models/) is
76paid for at those providers instead. Each such run here is a flat $0.10
77for g1t's orchestration, plus its [sandbox time](#sandbox-time).
78
79The charge goes to the workspace that owns the repository, whoever
80assigned the issue. That is why only members of a workspace can put g1t
81agents to work on its repositories.
82
83## How prices are set
84
85g1t passes its own costs through. Everything a workspace uses costs g1t
86money first, at Cloudflare or a model provider, and is charged at that
87cost plus a set markup. There is no seat price, and nothing is bundled to
88hide what it costs. The live prices are on
89[g1t.sh/pricing](https://g1t.sh/pricing), straight from the price book
90billing charges from.
91
92Prices keep themselves current as those costs move:
93
94- **Models.** Each of g1t's hosted runs goes through its Cloudflare AI
95 Gateway, which prices every request at the provider's current rates. A
96 run is charged when it finishes at what the sandbox reported; within
97 about 15 minutes it is **settled** to the gateway's figure, and any
98 difference appears on the statement as a correction, such as
99 *Correction to "Work on acme/api#12": AI Gateway priced its 41 model
100 requests at $0.0312, not $0.0298*. When a provider changes its prices,
101 runs are charged the new ones from that day. A run whose sandbox stopped
102 without reporting is charged from the gateway's logs instead of not at
103 all.
104- **Cloudflare.** Every day, g1t checks what Cloudflare billed its account
105 against what was used: Containers against the seconds containers ran,
106 Workers for Platforms per request and per CPU millisecond. When a cost
107 moves by 2% or more, the price book moves with it, since each price is
108 its cost times its markup, and the change is listed on the pricing page
109 with the reason. A measurement far from the current cost (more than 4×
110 either way) is not adopted, only logged, so one odd day cannot reprice
111 anything.
112
113| | Markup |
114| --- | --- |
115| Models | 20% |
116| Deploy builds, app requests, CPU and apps | 20% |
117| Sandbox time | 138%, which also pays for the orchestration around each sandbox and everyone's free minutes |
118
119## Sandbox time
120
121Every sandbox g1t starts for a workspace runs on Cloudflare Containers, and
122Cloudflare charges g1t for every second of it. So each one is metered by
123the second, from start to stop, whatever it was for: agents, reviews,
124revisions, catch-ups, planning, acceptance checks, the merge queue and
125workflow jobs. It is charged to the workspace that owns the repository.
126
127| | |
128| --- | --- |
129| Free each month | 500 minutes (calendar month, UTC) |
130| Past that | about $0.002 a minute, by the second |
131| What it costs g1t | about $0.0009 a minute (Containers, standard-1, at the CPU sandboxes really use) |
132
133Both follow what Cloudflare bills, so they move; today's exact figures are
134on [g1t.sh/pricing](https://g1t.sh/pricing). See
135[How prices are set](#how-prices-are-set).
136
137Deploy builds are not counted here: [Deployments](/guides/deployments/)
138charges them by the second on its own plan.
139
140Each sandbox is one line on the [statement](#the-statement), such as
141*Checks on acme/api#12: 3m 12s of sandbox time*, with whether it fell
142within the free minutes.
143
144## Usage limits
145
146Every workspace has two limits. One protects you from a surprise bill;
147the other protects g1t from usage that is never paid for. Both are on
148**Settings → Billing → Limits**.
149
150### Your monthly spend limit
151
152What the workspace may spend in a month (UTC). At it, work stops until the
153month turns or an owner raises it. Owners choose one of:
154
155- **Automatic** (the default): $200, or twice last month's spend,
156 whichever is more. It keeps up as you grow: a workspace that spent $900
157 last month can spend $1,800 this month without anyone changing a thing.
158- **Fixed**: an amount you set.
159- **None**: work never stops for spend.
160
161You are emailed at 50%, 80% and 100% of it.
162
163### What g1t lets go unpaid
164
165Usage is charged after it runs, so at any moment some of it is not yet
166paid for. g1t lets that reach a ceiling that grows with your history:
167
168| | Ceiling on what is unpaid |
169| --- | --- |
170| **New**: has not paid g1t yet | $3: the free allowances and a little more |
171| **Paid** | twice what you have paid, from $25 up to $1,000 |
172| **Established**: three steady months | three times your monthly spend, up to $10,000, by itself |
173| **Reviewed** | what g1t set with you; contact us |
174
175With a card on file, **g1t charges it as you near the ceiling** (80%):
176an invoice for what you owe, paid at once, after which the ceiling is
177yours again. So a workspace that pays keeps going, however much it uses;
178the ceiling only stops one that does not.
179
180How the ceiling grows:
181
182- A payment counts once it has **cleared for 7 days**, the time in which
183 most bad cards are caught. Payments with prepaid cards pay, but do not
184 raise the ceiling, nor do credits g1t gives or payments in test mode.
185- **Established** comes by itself after three months in a row of real spend
186 ($20 or more each), every monthly invoice paid, nothing declined in 90
187 days and nothing ever disputed. From then the ceiling follows your
188 spend.
189- Past $10,000, or for terms of your own, **contact us**: we set it with
190 you, often with an enterprise account and invoices.
191
192What counts as unpaid is each item at what it cost g1t or what it is
193charged, whichever is more, less what was paid this month, plus anything
194left unpaid from earlier months: a new month is not a fresh allowance.
195
196A **declined card** stops work until it is paid, as does a payment
197disputed with the card's bank. Paying under Billing with another card
198clears it at once.
199
200## Invoices
201
202Every charge is a real invoice from g1t, kept on Stripe's billing page
203with its PDF and emailed as a receipt:
204
205- **When each month closes**, an invoice for what the workspace owes,
206 itemised: agents on g1t's models, runs on your own model provider,
207 sandbox time, and deployments past the plan. Credit you paid in advance
208 is taken off as *Paid in advance*; anything left unpaid from before is
209 added.
210- **When the workspace nears its ceiling** mid-month, the same, sooner.
211
212Each is charged to the card on file. The Billing page lists them, with
213links to view each on Stripe and download its PDF.
214
215## Enterprises and custom terms
216
217Some accounts are billed differently, set up by g1t with you:
218
219- **Enterprise**: one billing account paying for several workspaces.
220 Their usage and payments count together, against
221 one limit, on one set of terms, and each workspace's Billing page says
222 which enterprise pays for it. An enterprise is invoiced: when each month
223 closes, Stripe emails one invoice to the enterprise's billing address,
224 with a line for each workspace, due in 30 days and paid on Stripe's
225 invoice page by card or bank transfer. Paying it clears every workspace
226 on it; if it goes overdue, their work stops until it is paid.
227- **Comped**: g1t covers the account's usage. Usage is still recorded with
228 what it cost, so the Usage page stays accurate, and paid features are on
229 without a plan.
230- **Custom**: a discount on every usage charge, a limit of its own, or
231 both, sometimes until a date, after which standard terms apply.
232
233Each change is made by g1t staff in g1t's billing console and recorded with
234who made it and why. To ask for one, write to support.
235
236## Your card, invoices and billing details
237
238These live on **Stripe's billing page**, not on g1t: g1t never sees or
239stores card numbers. An owner opens it from **Settings → Billing → Card
240and invoices → Manage billing on Stripe** (or **Add a card on Stripe**),
241and there adds or replaces the card, downloads invoices and receipts, and
242sets the billing email, address and tax ID. Saving a card charges
243nothing. g1t support never takes card details by phone or email; if you
244need help, we send you a link to that same Stripe page.
245
246With a card on file:
247
248- **Near the ceiling** on what is unpaid (80%), g1t sends an
249 [invoice](#invoices) for what the workspace owes and charges it, so work
250 does not stop.
251- **When each month closes**, the month's [invoice](#invoices) is charged
252 to it.
253- **If it is declined**, work stops until the workspace pays, and the
254 Billing page and the API say why. Replace the card or add credit to pay.
255
256Comped workspaces are never charged, and an enterprise's workspaces are
257billed through the enterprise. Test-mode cards are never charged
258automatically.
259
260## Add credit
261
262Credit is a payment in advance: it pays for usage as it happens, and lowers
263what the workspace owes against its limit. It is never needed to start
264work. Only an owner of the workspace can add credit.
265
2661. Open the workspace's **Settings → Billing**, `g1t.sh/<workspace>/-/billing`.
2672. Under **Add credit by card**, choose an amount: $10, $25, $50 or $100.
2683. Pay on the card page you are sent to.
269
270You come back to the Billing page, and the credit is there once the payment
271has gone through. The amount credited is what the card processor says was
272paid.
273
274While payments on g1t are in test mode, no real card is charged. Use the
275test card `4242 4242 4242 4242` with any future date and any code. The
276Billing page says when payments are in test mode.
277
278## When work is stopped
279
280A workspace at its limit, or with a declined card, starts nothing new.
281Assigning an issue, planning, or asking for a review is refused with `402`
282and the reason:
283
284```json
285{
286 "error": {
287 "code": "payment_required",
288 "message": "The acme workspace reached its $3.00 limit for usage not yet paid for, so its sandboxes, builds and apps are stopped. The limit grows as a workspace pays g1t; an owner can pay under Billing, or write to support to have it raised."
289 }
290}
291```
292
293A step g1t would take by itself, such as a revision or a review, stops
294instead, and the pull request says **Needs you** with the reason. Runs
295already under way finish, so usage can go slightly past the limit.
296
297## The Usage page
298
299A workspace's **Usage** page, `g1t.sh/<workspace>/-/usage`, shows what its
300agents have cost. Every member can see it. The sidebar shows this month's
301spend.
302
303While g1t is free, nothing is charged, so the page and the sidebar show
304what runs **used**, at what they cost: g1t's models and your own provider's
305together. Credit is not drawn down.
306
307These figures are what the agent harness reports for each run. Your model
308provider's or gateway's own figures can differ by a few percent, because
309each prices the same tokens itself; the provider's invoice is what counts.
310
311Pick a period: **This month**, **Last 7 days**, **Last 30 days** or **Last
31290 days**. The page then shows:
313
314| | |
315| --- | --- |
316| Spent | What the period cost, and how much of it was the model provider's. **Used**, at cost, while g1t is free. |
317| Agent runs | How many runs there were. |
318| Average run | What a run cost on average. |
319| Credit left | The balance, and about how many days it lasts at the period's rate. |
320| Spend per day | A chart of each day, split by kind of work. |
321| By kind of work | Making changes, reviews, catching up and planning. A revision counts as making a change. |
322| By repository | Each repository's share. |
323| Pull requests that cost most | The ten that cost most, each linked. Planning appears as the repository, linked to its plans. |
324| By model | Each model's share. |
325
326## The statement
327
328Beside the balance, the **Billing** page shows what the workspace has
329spent this month and in total, and what it has paid and been credited in
330total: every agent run, sandbox, build and app, at what was charged.
331
332The **Billing** page ends with the workspace's statement, a month at a
333time. Every member can see it.
334
335- **Totals.** What the month charged, what was paid and credited, and how
336 many entries make it up.
337- **Grouped by day or by project.** Each day (or project) has one line
338 per kind of charge, with how many entries it holds and what they come
339 to, rather than a row for every run:
340
341 | Line | What it holds |
342 | --- | --- |
343 | Agent runs | Runs on g1t's model provider: the model's cost plus the margin. |
344 | Runs on your own model provider | The run fee for runs billed by your own provider. |
345 | Sandbox time | Each sandbox's time, memory and disk. |
346 | Deployments | Builds and apps beyond the allowance. |
347 | Payments | Card payments and invoices paid. |
348 | Credits from g1t | Credit g1t added, such as a goodwill credit. |
349 | Refunds | Money given back to your card. |
350
351- **Open a line** to see its entries, 50 at a time, newest first. Each
352 run links to the pull request it was for.
353- **Pick a month** to see an earlier one; months with no entries are not
354 listed.
355- **CSV** downloads every entry of the month, with the line it falls
356 under, for your own books. Invoices from Stripe remain the record for
357 what was charged to your card.
358
359Each pull request's session also ends with what its run cost before the
360margin.
361
362## The preview
363
364g1t is in preview.
365
366- **Open to everyone:** accounts, workspaces, repositories, git, issues,
367 pull requests, review, the API, and your own agent through MCP.
368- **g1t's agents, for any workspace with its own model provider:** connect
369 an Anthropic key or endpoint under [Integrations](/guides/models/) and the
370 workspace's agents, acceptance checks and merge queue work at once. Your
371 provider bills you for the models; g1t charges nothing for now, and $0.10
372 a run once pricing starts.
373- **g1t's hosted models:** while payments are in test mode, every
374 workspace can use g1t's own models on
375 [the free allowance](#the-free-allowance) ($1 each, until October 22),
376 and a few g1t has opened them to without limit. When payments go live,
377 every workspace can use them, paid from its credit.
378
379Each workspace decides where its model spend goes. A workspace that can use
380neither sees a message saying so, with the way to connect its own provider.