AI Gateway: your own code calls models with a workspace token · 36a518a · flagon-io/g1t

269	269		token_id: "tok_1".to_owned(),
270	270		scopes: scopes.map(\|scopes\| scopes.iter().map(\|scope\| scope.as_str().to_owned()).collect()),
271	271		legacy,
	272	+	name: None,
272	273		})),
273	274		..User::default()
274	275		}

412	412		})
413	413		}
414	414		Op::GetBillingDetails => shaped(g1t_kit::call(billing, "billing_details", &account).await?, &details_json),
	415	+	Op::ListGatewayRequests => {
	416	+	let limit = match &input["limit"] {
	417	+	Value::Null => None,
	418	+	Value::Number(number) => number.as_u64(),
	419	+	Value::String(digits) => digits.trim().parse().ok(),
	420	+	_ => Some(0),
	421	+	};
	422	+	if limit.is_some_and(\|limit\| !(1..=200).contains(&limit)) {
	423	+	return failed(FailureCode::Invalid, "limit is a number from 1 to 200.");
	424	+	}
	425	+	let before = input["before"].as_str().map(str::trim).filter(\|id\| !id.is_empty());
	426	+	let page: Outcome<Value> = g1t_kit::call(
	427	+	billing,
	428	+	"gateway_requests",
	429	+	&json!({ "workspace": workspace, "viewer": viewer, "limit": limit, "before": before }),
	430	+	)
	431	+	.await?;
	432	+	shaped(page, &snake)
	433	+	}
415	434		_ => failed(FailureCode::Invalid, "Not a billing operation."),
416	435		}
417	436		}
⋯
559	578		assert!(shown.get("invoices").is_none() && shown.get("upcoming").is_none());
560	579		}
561	580
562		−	const OPS: [Op; 7] =
563		−	[Op::GetUsage, Op::GetBudget, Op::SetBudget, Op::GetAiCredit, Op::BuyAiCredit, Op::ListInvoices, Op::GetBillingDetails];
	581	+	const OPS: [Op; 8] = [
	582	+	Op::GetUsage,
	583	+	Op::GetBudget,
	584	+	Op::SetBudget,
	585	+	Op::GetAiCredit,
	586	+	Op::BuyAiCredit,
	587	+	Op::ListInvoices,
	588	+	Op::GetBillingDetails,
	589	+	Op::ListGatewayRequests,
	590	+	];
564	591
565	592		/// Billing belongs to a workspace, needs someone signed in, and is one
566	593		/// MCP tool whose writes no preset but full access reaches.
⋯
579	606		token_id: "tok_1".into(),
580	607		scopes: preset.scopes().map(\|scopes\| scopes.iter().map(\|scope\| scope.as_str().to_owned()).collect()),
581	608		legacy: false,
	609	+	name: None,
582	610		};
583	611		for preset in [Preset::ReadOnly, Preset::Agent] {
584	612		let access = token(preset);
585	613		let seen: Vec<&str> = tool.visible(&Gate::Token(&access)).iter().map(\|action\| action.name).collect();
586		−	assert_eq!(seen, ["usage", "budget", "ai_credit", "invoices", "billing_details"], "{}", preset.as_str());
	614	+	assert_eq!(seen, ["usage", "budget", "ai_credit", "invoices", "billing_details", "gateway_requests"], "{}", preset.as_str());
587	615		}
	616	+	// The AI Gateway's log needs models:read, and nothing of billing's.
	617	+	let models = TokenAccess { scopes: Some(vec!["models:read".into()]), ..token(Preset::Ci) };
	618	+	let seen: Vec<&str> = tool.visible(&Gate::Token(&models)).iter().map(\|action\| action.name).collect();
	619	+	assert_eq!(seen, ["gateway_requests"]);
588	620		let full = TokenAccess::full();
589	621		assert_eq!(tool.visible(&Gate::Token(&full)).len(), OPS.len());
590	622		}

63	63		),
64	64		(
65	65		"Billing",
66		−	"A workspace's usage, its budget, its AI credit and its invoices. Members read them; owners change the budget and buy credit, as people. g1t's agents never change billing.",
	66	+	"A workspace's usage, its budget, its AI credit, its invoices and its AI Gateway requests. Members read them; owners change the budget and buy credit, as people. g1t's agents never change billing.",
67	67		&[
68	68		Op::GetUsage,
69	69		Op::GetBudget,
⋯
72	72		Op::BuyAiCredit,
73	73		Op::ListInvoices,
74	74		Op::GetBillingDetails,
	75	+	Op::ListGatewayRequests,
75	76		],
76	77		),
77	78		(
⋯
513	514		Op::BuyAiCredit => "Buy AI credit",
514	515		Op::ListInvoices => "List a workspace's invoices",
515	516		Op::GetBillingDetails => "Get a workspace's billing details",
	517	+	Op::ListGatewayRequests => "List a workspace's AI Gateway requests",
516	518		Op::PinProject => "Pin a project",
517	519		Op::UnpinProject => "Unpin a project",
518	520		Op::ReorderPinnedProjects => "Reorder your pinned projects",

258	258		BuyAiCredit,
259	259		ListInvoices,
260	260		GetBillingDetails,
	261	+	ListGatewayRequests,
261	262		RequestReviewers,
262	263		RemoveRequestedReviewers,
263	264		GetCodeownersErrors,
⋯
625	626		}
626	627
627	628		impl Op {
628		−	pub const ALL: [Op; 205] = [
	629	+	pub const ALL: [Op; 206] = [
629	630		Op::Whoami,
630	631		Op::GetWorkspace,
631	632		Op::CreateWorkspace,
⋯
797	798		Op::BuyAiCredit,
798	799		Op::ListInvoices,
799	800		Op::GetBillingDetails,
	801	+	Op::ListGatewayRequests,
800	802		Op::RequestReviewers,
801	803		Op::RemoveRequestedReviewers,
802	804		Op::GetCodeownersErrors,
⋯
1011	1013		Op::BuyAiCredit => "buy_ai_credit",
1012	1014		Op::ListInvoices => "list_invoices",
1013	1015		Op::GetBillingDetails => "get_billing_details",
	1016	+	Op::ListGatewayRequests => "list_gateway_requests",
1014	1017		Op::RequestReviewers => "request_reviewers",
1015	1018		Op::RemoveRequestedReviewers => "remove_requested_reviewers",
1016	1019		Op::GetCodeownersErrors => "get_codeowners_errors",
⋯
1496	1499		Op::GetBillingDetails => {
1497	1500		"Who a workspace's invoices are made out to: the billing `email`, `name`, `address`, tax ID (`tax_id_type`, `tax_id`), `po_number` and the invoices' `language`, with the default `payment_method` as far as it is safe to show (its kind, brand, last four digits and expiry). `customer` is false until the workspace has been set up to pay. Tax is worked out from the address: `tax_location` says whether it is enough for that (a country, and in the US a ZIP code), `tax_address_needed_at` is set while g1t is holding a charge for want of one, `tax_id_status` is Stripe's check of the tax ID (`pending`, `verified`, `unverified` or `unavailable`), and `tax_exempt` is `none`, `exempt` or `reverse`. Members of the workspace only."
1498	1501		}
	1502	+	Op::ListGatewayRequests => {
	1503	+	"A workspace's recent AI Gateway requests, newest first: each with its `id`, `created_at`, `model`, the access token that sent it (`token_id`, `token_name`), its tokens by kind (`input`, `output`, `cache_read`, `cache_write`), what they cost at the model's price (`cost_micros`) and what the workspace was charged (`charged_micros`, before included usage and AI credit paid for it; 0 on the workspace's own provider key, `own_key`), the HTTP `status` it was answered with, whether it was `streamed`, `duration_ms`, and `error` for one that was refused or failed. Prompts and answers are never kept. `limit` is how many, 50 unless given and 200 at most; pass `next` from one page as `before` for the next. Requests are kept `retention_days` (30). Members of the workspace only."
	1504	+	}
1499	1505		Op::ListUserTeams => {
1500	1506		"The teams someone is in within a workspace, as list_teams describes them, leaving out secret teams you cannot see. Members of the workspace only."
1501	1507		}
⋯
2754	2760		Op::GetBudget \| Op::GetAiCredit \| Op::ListInvoices \| Op::GetBillingDetails => {
2755	2761		object(json!({ "workspace": workspace_schema() }), &["workspace"])
2756	2762		}
	2763	+	Op::ListGatewayRequests => object(
	2764	+	json!({
	2765	+	"workspace": workspace_schema(),
	2766	+	"limit": { "type": "integer", "minimum": 1, "maximum": 200, "description": "How many requests, newest first. 50 if not given." },
	2767	+	"before": { "type": "string", "description": "Only requests older than this one: the `next` of the page before." },
	2768	+	}),
	2769	+	&["workspace"],
	2770	+	),
2757	2771		Op::SetBudget => object(
2758	2772		json!({
2759	2773		"workspace": workspace_schema(),
⋯
2936	2950		\| Op::BuyAiCredit
2937	2951		\| Op::ListInvoices
2938	2952		\| Op::GetBillingDetails
	2953	+	\| Op::ListGatewayRequests
2939	2954		)
2940	2955		}
2941	2956
⋯
4729	4744		\| Op::GetAiCredit
4730	4745		\| Op::BuyAiCredit
4731	4746		\| Op::ListInvoices
4732		−	\| Op::GetBillingDetails => crate::billing::run(self, services, viewer, input).await,
	4747	+	\| Op::GetBillingDetails
	4748	+	\| Op::ListGatewayRequests => crate::billing::run(self, services, viewer, input).await,
4733	4749		Op::ListUserTeams => {
4734	4750		pass(
4735	4751		identity,

5961	5961		},
5962	5962		"notes": "Open `url` in a browser to pay. Nothing is charged until the payment is made there; the credit then shows in get_ai_credit. An amount outside `min_cents` to `max_cents`, or not in whole dollars, is refused with `invalid`; a workspace that cannot buy credit (see `can_buy`) gets `conflict` or `payment_required`."
5963	5963		},
	5964	+	"list_gateway_requests": {
	5965	+	"params": {
	5966	+	"workspace": "flagon-io"
	5967	+	},
	5968	+	"query": {
	5969	+	"limit": 2
	5970	+	},
	5971	+	"response": {
	5972	+	"requests": [
	5973	+	{
	5974	+	"id": "gw_5f0c2a9e7b1d4c3a8e6f0b12",
	5975	+	"created_at": "2026-10-07T14:02:11.000Z",
	5976	+	"model": "claude-sonnet-5-5",
	5977	+	"token_id": "tok_01kkr4w7d2c9e5f8h1j3m6n0p2",
	5978	+	"token_name": "release-notes",
	5979	+	"input": 1840,
	5980	+	"output": 512,
	5981	+	"cache_read": 12000,
	5982	+	"cache_write": 0,
	5983	+	"cost_micros": 11200,
	5984	+	"charged_micros": 11200,
	5985	+	"status": 200,
	5986	+	"own_key": false,
	5987	+	"streamed": true,
	5988	+	"duration_ms": 4210,
	5989	+	"error": null
	5990	+	},
	5991	+	{
	5992	+	"id": "gw_1a7e3c5b9d2f4e6a8c0b2d41",
	5993	+	"created_at": "2026-10-07T13:58:40.000Z",
	5994	+	"model": "claude-opus-5-5",
	5995	+	"token_id": "tok_01kkr4w7d2c9e5f8h1j3m6n0p2",
	5996	+	"token_name": "release-notes",
	5997	+	"input": 0,
	5998	+	"output": 0,
	5999	+	"cache_read": 0,
	6000	+	"cache_write": 0,
	6001	+	"cost_micros": 0,
	6002	+	"charged_micros": 0,
	6003	+	"status": 402,
	6004	+	"own_key": false,
	6005	+	"streamed": false,
	6006	+	"duration_ms": 38,
	6007	+	"error": "The flagon-io workspace is out of AI credit and has used this month's included usage, so the AI Gateway refuses requests to g1t's models. An owner can buy AI credit or turn on auto-reload at /flagon-io/-/billing#ai-credit."
	6008	+	}
	6009	+	],
	6010	+	"next": "gw_1a7e3c5b9d2f4e6a8c0b2d41",
	6011	+	"retention_days": 30
	6012	+	},
	6013	+	"notes": "To send requests, see the AI Gateway guide: POST https://models.g1t.sh/anthropic/v1/messages with a workspace access token that has `models:write`. `charged_micros` is what the request was charged before included usage and AI credit paid for it; the payment itself is on the statement as AI Gateway. Pass `next` as `before` for the next page; it is null on the last."
	6014	+	},
5964	6015		"list_invoices": {
5965	6016		"params": {
5966	6017		"workspace": "flagon-io"

134	134		route("POST", "/workspaces/:workspace/ai_credit/checkout", Op::BuyAiCredit, &[]),
135	135		route("GET", "/workspaces/:workspace/invoices", Op::ListInvoices, &[]),
136	136		route("GET", "/workspaces/:workspace/billing_details", Op::GetBillingDetails, &[]),
	137	+	// The AI Gateway's log of a workspace's requests.
	138	+	route("GET", "/workspaces/:workspace/gateway/requests", Op::ListGatewayRequests, &[("limit", "limit"), ("before", "before")]),
137	139		// Code owners: the CODEOWNERS file, checked.
138	140		route("GET", "/repos/:owner/:name/codeowners/errors", Op::GetCodeownersErrors, &[("ref", "ref")]),
139	141		// Security alerts: secrets and vulnerable dependencies.
⋯
1031	1033		assert_eq!(op("POST", "/workspaces/acme/ai_credit/checkout"), Op::BuyAiCredit);
1032	1034		assert_eq!(op("GET", "/workspaces/acme/invoices"), Op::ListInvoices);
1033	1035		assert_eq!(op("GET", "/workspaces/acme/billing_details"), Op::GetBillingDetails);
	1036	+	assert_eq!(op("GET", "/workspaces/acme/gateway/requests"), Op::ListGatewayRequests);
1034	1037		}
1035	1038
1036	1039		#[test]

292	292		Tool {
293	293		name: "billing",
294	294		title: "Billing",
295		−	description: "A workspace's billing: its usage by product, project and day, its budget (the monthly spend limit, alerts and whether usage pauses at it), its AI credit, and its invoices. Amounts are whole millionths of a dollar (`_micros`), or cents (`_cents`) where named. Members read it; changing the budget and buying credit are for owners, as people, and never for g1t's agents.",
	295	+	description: "A workspace's billing: its usage by product, project and day, its budget (the monthly spend limit, alerts and whether usage pauses at it), its AI credit, its invoices, and its AI Gateway requests. Amounts are whole millionths of a dollar (`_micros`), or cents (`_cents`) where named. Members read it; changing the budget and buying credit are for owners, as people, and never for g1t's agents.",
296	296		default_action: Some("usage"),
297	297		actions: &[
298	298		a("usage", Op::GetUsage, "Usage over a range of days, by product, meter, project and day, and what paid for it"),
⋯
302	302		a("buy_ai_credit", Op::BuyAiCredit, "A payment page to buy AI credit, for a person to open"),
303	303		a("invoices", Op::ListInvoices, "Every invoice, the itemised usage invoices, and the next one so far"),
304	304		a("billing_details", Op::GetBillingDetails, "Who invoices are made out to and the payment method on file"),
	305	+	a("gateway_requests", Op::ListGatewayRequests, "Recent AI Gateway requests: model, tokens, cost, status and token"),
305	306		],
306	307		},
307	308		Tool {
⋯
640	641		token_id: "tok_1".to_owned(),
641	642		scopes: scopes.map(\|scopes\| scopes.iter().map(\|scope\| scope.as_str().to_owned()).collect()),
642	643		legacy: false,
	644	+	name: None,
643	645		}
644	646		}
645	647

115	115		items: [
116	116		{ label: 'Integrations', slug: 'guides/integrations' },
117	117		{ label: 'Model providers', slug: 'guides/models' },
	118	+	{ label: 'AI Gateway', slug: 'guides/ai-gateway' },
118	119		{ label: 'Webhooks', slug: 'guides/webhooks' },
119	120		{ label: 'GitHub Actions', slug: 'guides/actions' },
120	121		{ label: 'Self-hosted runners', slug: 'guides/self-hosted-runners' },

1	+	---
2	+	title: AI Gateway
3	+	description: Send your own code's model requests through g1t with a workspace access token, paid from AI credit at the model's price, with a log of every request.
4	+	---
5	+
6	+	The AI Gateway takes model requests from your own code, in Anthropic's
7	+	Messages format, and sends them to the model. You point an Anthropic SDK,
8	+	Claude Code or anything else that speaks that format at one base URL and
9	+	give it a workspace access token as its API key:
10	+
11	+	\| \| \|
12	+	\| --- \| --- \|
13	+	\| Base URL \| `https://models.g1t.sh/anthropic` \|
14	+	\| API key \| A workspace access token (`g1t_…`) with the `models:write` scope \|
15	+
16	+	Each request is charged to the workspace at the model's price and paid from
17	+	the plan's included usage and [AI credit](/guides/usage-and-billing/#ai-credit).
18	+	While the gateway is in beta there is no markup. If the workspace has
19	+	connected its own Anthropic key, requests go there instead and cost
20	+	nothing on g1t. Every request is logged with its model, tokens, cost and
21	+	status. Prompts and answers are never kept.
22	+
23	+	## Before you start
24	+
25	+	You need one of these:
26	+
27	+	- The workspace on the g1t plan, with AI credit or this month's
28	+	included usage left. See [AI credit](/guides/usage-and-billing/#ai-credit).
29	+	- The workspace's own Anthropic key, connected under
30	+	[Integrations](/guides/models/#connect-a-provider). Then the plan is not
31	+	needed and nothing is charged.
32	+
33	+	## Make a token
34	+
35	+	The gateway takes a workspace's own token, so its usage is the workspace's
36	+	and keeps working when the person who set it up leaves. Only owners make
37	+	them.
38	+
39	+	1. Open the workspace's Settings → Access tokens.
40	+	2. Under New token, give it a name, such as `release-notes`. The log
41	+	shows each request's token by this name.
42	+	3. The scopes start on the CI preset. Untick what the code does not need,
43	+	and under AI Gateway tick `models:write`. A token with only
44	+	`models:write` can send model requests and nothing else.
45	+	4. Choose an expiry and select Create token. Copy the token now: it is
46	+	not shown again.
47	+
48	+	A personal access token is refused, even with `models:write`: the gateway
49	+	has to know which workspace to charge. A token with full access has every
50	+	scope, `models:write` included. See [scopes](/guides/authentication/#scopes).
51	+
52	+	## Send a request
53	+
54	+	The gateway answers the same routes as Anthropic's API, below the base URL:
55	+
56	+	\| Route \| What it does \|
57	+	\| --- \| --- \|
58	+	\| `POST /anthropic/v1/messages` \| A message, streamed (`"stream": true`) or whole. Logged and charged. \|
59	+	\| `POST /anthropic/v1/messages/count_tokens` \| Counts a request's input tokens. Not logged, and costs nothing. \|
60	+
61	+	The token goes in `x-api-key`, or in `Authorization: Bearer`. Request and
62	+	answer bodies are Anthropic's, unchanged, and so are streamed events.
63	+
64	+	With curl:
65	+
66	+	```sh
67	+	curl https://models.g1t.sh/anthropic/v1/messages \
68	+	-H "x-api-key: $G1T_TOKEN" \
69	+	-H "anthropic-version: 2023-06-01" \
70	+	-H "content-type: application/json" \
71	+	-d '{
72	+	"model": "claude-sonnet-5-5",
73	+	"max_tokens": 1024,
74	+	"messages": [{ "role": "user", "content": "Write a commit message for: fix the login redirect" }]
75	+	}'
76	+	```
77	+
78	+	With Anthropic's TypeScript SDK:
79	+
80	+	```ts
81	+	import Anthropic from "@anthropic-ai/sdk";
82	+
83	+	const client = new Anthropic({
84	+	baseURL: "https://models.g1t.sh/anthropic",
85	+	apiKey: process.env.G1T_TOKEN,
86	+	});
87	+
88	+	const message = await client.messages.create({
89	+	model: "claude-sonnet-5-5",
90	+	max_tokens: 1024,
91	+	messages: [{ role: "user", content: "Write a commit message for: fix the login redirect" }],
92	+	});
93	+	```
94	+
95	+	With Anthropic's Python SDK:
96	+
97	+	```python
98	+	import os
99	+
100	+	import anthropic
101	+
102	+	client = anthropic.Anthropic(
103	+	base_url="https://models.g1t.sh/anthropic",
104	+	api_key=os.environ["G1T_TOKEN"],
105	+	)
106	+
107	+	message = client.messages.create(
108	+	model="claude-sonnet-5-5",
109	+	max_tokens=1024,
110	+	messages=[{"role": "user", "content": "Write a commit message for: fix the login redirect"}],
111	+	)
112	+	```
113	+
114	+	### Claude Code
115	+
116	+	Set two environment variables before you start it:
117	+
118	+	```sh
119	+	export ANTHROPIC_BASE_URL=https://models.g1t.sh/anthropic
120	+	export ANTHROPIC_AUTH_TOKEN=g1t_…
121	+	claude
122	+	```
123	+
124	+	Claude Code's own small requests go to Claude Haiku 4.5, which the gateway
125	+	offers. To choose the main model, also set `ANTHROPIC_MODEL`, such as
126	+	`claude-opus-5-5`.
127	+
128	+	## Models
129	+
130	+	On g1t's models the gateway offers these. Prices are per million tokens,
131	+	the provider's list price; cache writes are five-minute ones.
132	+
133	+	\| Model \| `model` \| Input \| Output \| Cache reads \| Cache writes \|
134	+	\| --- \| --- \| --- \| --- \| --- \| --- \|
135	+	\| Claude Opus 5.5 \| `claude-opus-5-5` \| $4.00 \| $20.00 \| $0.20 \| $5.00 \|
136	+	\| Claude Sonnet 5.5 \| `claude-sonnet-5-5` \| $2.00 \| $10.00 \| $0.20 \| $2.50 \|
137	+	\| Claude Haiku 4.5 \| `claude-haiku-4-5`, `claude-haiku-4-5-20251001` \| $1.00 \| $5.00 \| $0.10 \| $1.25 \|
138	+
139	+	A request for any other model is refused with `400` before it reaches the
140	+	provider, and the error names the models offered. On the workspace's own
141	+	key, a request can name any model that key can use.
142	+
143	+	On g1t's models a request is charged only by its tokens, so what the
144	+	provider bills some other way is refused with `400` for now:
145	+
146	+	\| Not offered on g1t's models yet \| In the request \|
147	+	\| --- \| --- \|
148	+	\| Fast mode \| `speed` other than `standard` \|
149	+	\| Inference in one region \| `inference_geo` other than `global` \|
150	+	\| Server-side fallbacks \| `fallbacks` \|
151	+	\| Server tools, such as web search, web fetch and code execution \| A tool whose `type` is not your own (`custom` or none) or a client tool (`bash_…`, `text_editor_…`, `computer_…`, `memory_…`) \|
152	+	\| Containers and skills \| `container` \|
153	+
154	+	All of them work on the workspace's own key, which the provider bills. In
155	+	Claude Code on g1t's models, its web search fails for this reason; the
156	+	rest of Claude Code works.
157	+
158	+	Anthropic's format is the one served today. OpenAI's format and open
159	+	models are coming later.
160	+
161	+	## What it costs
162	+
163	+	\| Where it goes \| You pay \|
164	+	\| --- \| --- \|
165	+	\| g1t's models \| Its tokens at the model's price above, with no markup while the gateway is in beta \|
166	+	\| The workspace's own Anthropic key \| Nothing on g1t. The provider bills you for the model. \|
167	+
168	+	On g1t's models:
169	+
170	+	- Each request that used tokens is one line on the statement, under
171	+	AI Gateway, such as *AI Gateway: Claude Sonnet 5.5, 14,352 tokens,
172	+	token release-notes*.
173	+	- The plan's included usage pays first, then AI credit. Trial credit and
174	+	g1t's open-source pool never pay for gateway requests.
175	+	- It is not an agent run, so the [agent rate](/guides/usage-and-billing/#the-agent-rate)
176	+	does not apply.
177	+	- It counts toward the workspace's [spend limit](/guides/usage-and-billing/#your-spend-limit)
178	+	like any other usage, and shows on Usage under the AI Gateway product.
179	+	- A workspace with a 100% discount gets it free through the discount; an
180	+	enterprise is invoiced for it after use.
181	+
182	+	### Your own key
183	+
184	+	When the workspace has an Anthropic or Anthropic-compatible model provider
185	+	under [Integrations](/guides/models/), the gateway sends every request to
186	+	the first one connected, with its key. Those requests are logged with their
187	+	tokens and marked Own key, and g1t charges nothing for them. Remove the
188	+	provider and requests go to g1t's models again within a few seconds.
189	+
190	+	## Limits and errors
191	+
192	+	A request on g1t's models is refused before it reaches the model when:
193	+
194	+	- The workspace is over its spend limit.
195	+	- It is on the plan, and has no AI credit and none of this month's included
196	+	usage left. If auto-reload is on, g1t tries it first.
197	+	- It is not on the g1t plan.
198	+
199	+	Errors are Anthropic's shape, so SDKs raise their usual errors:
200	+
201	+	```json
202	+	{ "type": "error", "error": { "type": "billing_error", "message": "The acme workspace is out of AI credit …" } }
203	+	```
204	+
205	+	\| Status \| `error.type` \| Why \|
206	+	\| --- \| --- \| --- \|
207	+	\| `400` \| `invalid_request_error` \| The body is not JSON, the model is not offered, or the request asks for something [not offered on g1t's models yet](#models). \|
208	+	\| `401` \| `authentication_error` \| The token is unknown, expired or deleted. \|
209	+	\| `402` \| `billing_error` \| Out of AI credit, over the spend limit, or not on the plan. The message says what an owner can do. \|
210	+	\| `403` \| `permission_error` \| Not a workspace's token, or it lacks `models:write`. \|
211	+	\| `404` \| `not_found_error` \| A route the gateway does not answer. \|
212	+
213	+	An error from the model provider, such as `429` or `529`, comes back as
214	+	the provider sent it. Refused and failed requests are logged with their
215	+	status and why, and cost nothing.
216	+
217	+	A deleted token, a provider added under Integrations, or AI credit just
218	+	bought takes effect within about ten seconds.
219	+
220	+	## See every request
221	+
222	+	The AI Gateway page lists the workspace's requests, newest first. Open
223	+	it from the link under Usage, or at `g1t.sh/<workspace>/-/gateway`.
224	+	Every member can see it.
225	+
226	+	\| Column \| \|
227	+	\| --- \| --- \|
228	+	\| Time \| When it was sent. Hover for the exact time, how long it took and whether it streamed. \|
229	+	\| Model \| The model it named. On the workspace's own key, the one that answered. \|
230	+	\| Input, Output, Cache read, Cache write \| Its tokens by kind. \|
231	+	\| Cost \| What it was charged, before included usage and AI credit paid for it, or Own key. \|
232	+	\| Status \| The status it was answered with. Hover a refusal or failure for why. \|
233	+	\| Token \| The name of the token that sent it. \|
234	+
235	+	Requests are kept 30 days.
236	+
237	+	From code, list them with
238	+	[`GET /workspaces/{workspace}/gateway/requests`](/reference/api/billing/list-gateway-requests/),
239	+	or the `billing` MCP tool's
240	+	[`gateway_requests`](/reference/mcp/#billing) action. Both need
241	+	`models:read`, which the Read only and Agent presets include.

328	328		\| Workspace \| `workspace:read`, `access:read`, `webhooks:read`, `secrets:read` \|
329	329		\| Billing \| `billing:read`, `billing:write` \|
330	330		\| Runners \| `runners:read` \|
	331	+	\| AI Gateway \| `models:read`, `models:write` \|
331	332		\| Dangerous \| `repo:admin`, `packages:delete`, `workspace:admin`, `access:admin`, `webhooks:admin`, `secrets:admin`, `runners:admin` \|
332	333
333	334		Ticking a higher level ticks the lower ones of its resource and greys
⋯
371	372		\| `secrets:admin` \| Set and delete secrets and variables \|
372	373		\| `runners:read` \| See [self-hosted runners](/guides/self-hosted-runners/), their groups and where agents run. Not in the Agent preset. \|
373	374		\| `runners:admin` \| Register and remove self-hosted runners, change their groups and settings \|
	375	+	\| `models:read` \| See the workspace's [AI Gateway](/guides/ai-gateway/) requests: their models, tokens, cost and status \|
	376	+	\| `models:write` \| Send model requests through the [AI Gateway](/guides/ai-gateway/), which uses the workspace's AI credit. Only a workspace's own token can send them. Not in any preset but full access. \|
374	377
375	378		Every operation of the API and the MCP server needs exactly one of these,
376	379		except `whoami` (`GET /user`), which any token may use. Each endpoint's page

18	18		kind of work, which provider and model it runs on. Each provider bills
19	19		you for the model directly. Open to every workspace now.
20	20
	21	+	To call models from your own code with a workspace token, paid from the
	22	+	same AI credit, use the [AI Gateway](/guides/ai-gateway/).
	23	+
21	24		## Auto
22	25
23	26		On g1t's models you do not have to pick a model. Auto, the default,

123	123		\| Agent models \| A run \| What the provider charged \| The provider's price, from [AI credit](#ai-credit) \|
124	124		\| g1t agent rate \| Million tokens a run uses (input, output and cached), [weighted by kind](#the-agent-rate) \| — \| $0.25, from Oct 22, 2026 \|
125	125		\| g1t agent rate, your own model key \| The same, on runs that use [your own provider](/guides/models/) \| — \| $0.25, from Oct 22, 2026 \|
126		−	\| AI Gateway \| A request \| What the provider charged \| The provider's price: free of markup during beta \|
	126	+	\| [AI Gateway](/guides/ai-gateway/) \| A request \| What the provider charged \| The model's list price per token: free of markup during beta \|
127	127		\| Sandbox time (agents, workflows, the merge queue) \| Second \| About $0.001 a minute \| About $0.0012 a minute \|
128	128		\| [Larger machines](#workflow-jobs-on-larger-machines) for workflow jobs (`g1t-2core`, `g1t-4core`) \| Second \| About 2.8 and 5.1 times a sandbox second \| Cost + 20% \|
129	129		\| Deploy builds \| Second \| About $0.001 a minute \| About $0.0012 a minute \|

63	63		gets a [pull request that upgrades it](/guides/security/).
64	64		</Card>
65	65		<Card title="Your models, your tools" icon="plug">
66		−	Use g1t's models or [your own providers](/guides/models/), and pull context
	66	+	Use g1t's models or [your own providers](/guides/models/), call the same
	67	+	models from your own code through the [AI Gateway](/guides/ai-gateway/), and pull context
67	68		from [Sentry, Jira and Linear](/guides/integrations/).
68	69		</Card>
69	70		<Card title="Everything has an API" icon="book-open">

197	197		\| [Accounts](/reference/api/accounts/whoami/) \| Signing in from a tool, and who a token acts as. \|
198	198		\| [Workspaces](/reference/api/workspaces/create-workspace/) \| Creating a workspace. \|
199	199		\| [Notifications](/reference/api/notifications/list-notifications/) \| Your inbox: its threads, why you were told of each, marking them read, done, saved or snoozed, and what you subscribe to and watch. See [your inbox](/guides/inbox/). \|
200		−	\| [Billing](/reference/api/billing/get-usage/) \| A workspace's usage by product, project and day, its budget, its AI credit and its invoices. See [usage and billing](/guides/usage-and-billing/). \|
	200	+	\| [Billing](/reference/api/billing/get-usage/) \| A workspace's usage by product, project and day, its budget, its AI credit, its invoices and its [AI Gateway](/guides/ai-gateway/) requests. See [usage and billing](/guides/usage-and-billing/). \|
201	201		\| [Invites](/reference/api/invites/list-invites/) \| Your invites while g1t is invite-only, and inviting people into a workspace by email. \|
202	202		\| [Repositories](/reference/api/repositories/list-repos/) \| A repository, how it handles pull requests, and its timeline. \|
203	203		\| [Access](/reference/api/access/list-collaborators/) \| Who has which role on a repository, invitations, outside collaborators, and a workspace's base permission. \|

552	552
553	553		## `billing`
554	554
555		−	A workspace's billing: its usage, its budget, its AI credit and its
556		−	invoices. `usage` is the default action. Amounts are whole millionths of a
	555	+	A workspace's billing: its usage, its budget, its AI credit, its
	556	+	invoices and its [AI Gateway](/guides/ai-gateway/) requests. `usage` is the default action. Amounts are whole millionths of a
557	557		dollar (`_micros`), or cents where a field says `_cents`. Members of the
558	558		workspace read it, a workspace's own token included. Changing the budget
559	559		and buying AI credit are for its owners, as people: signed in or with a
⋯
570	570		\| [`buy_ai_credit`](/reference/api/billing/buy-ai-credit/) \| A payment page (`url`) to buy `amount_cents` of credit, in whole dollars from $10 to $1,000, for a person to open and pay; it returns to the workspace's billing page. Owners, as people. \| `workspace`, `amount_cents` \| `billing:write` \|
571	571		\| [`invoices`](/reference/api/billing/list-invoices/) \| Every invoice (`invoices`, in cents), g1t's itemised usage invoices (`usage_invoices`), and what the next one comes to so far (`upcoming`). \| `workspace` \| `billing:read` \|
572	572		\| [`billing_details`](/reference/api/billing/get-billing-details/) \| Who invoices are made out to, and the payment method on file as far as it is safe to show. \| `workspace` \| `billing:read` \|
	573	+	\| [`gateway_requests`](/reference/api/billing/list-gateway-requests/) \| The workspace's recent [AI Gateway](/guides/ai-gateway/) requests, newest first: model, tokens by kind, `cost_micros`, `charged_micros`, `status`, `own_key` and the token that sent each. `limit` (50, at most 200) and `before` (the last page's `next`) page through them. Kept 30 days. \| `workspace` \| `models:read` \|
573	574
574	575		## `notifications`
575	576

Binary or large file; its contents are not shown.