Guide

Grok as a Janitor AI Proxy: The xAI Endpoint, What a Message Costs, and the EU Delay to Read First

Grok as a Janitor AI proxy: the OpenAI-compatible xAI endpoint, pricing per message, the OpenRouter route, and the EU caveats.

HProxy Team··5 min read
HProxy.Guide

Skip the dead lists.

Our free proxy list re-checks every exit every few minutes across 100+ countries, with a live last-checked time, so you copy IPs that worked moments ago, not a stale text dump.

Open the free proxy list→

Grok is the model Janitor AI users reach for when they want xAI's voice, and it connects more easily than its reputation suggests: xAI's API is OpenAI-compatible, so Grok goes into Janitor's proxy field the same way OpenRouter or DeepSeek does. It is not free in the API, the free Grok lives in the X app, and the content is governed by a policy stricter than the uncensored models most roleplayers default to. This guide gives the exact values, the per-message cost at xAI's listed prices, and the two caveats, content and region, that a proxy meets but does not dissolve.

The direct route: xAI's API

xAI documents its API as usable through the OpenAI SDK by pointing the base URL at https://api.x.ai/v1, and its API reference lists POST /v1/chat/completions. xAI now says "Chat Completions is offered as a legacy endpoint": it is still served, and new features reach its Responses API first. The Janitor proxy field speaks Chat Completions, so that is the address it wants.

FieldValue
Config NameGrok
Proxy URLhttps://api.x.ai/v1/chat/completions
API Keya key from console.x.ai
Modelgrok-4.7 (or grok-4.6, grok-4.3)

Menu, API Settings, Proxy, Add Configuration, fill in, save, refresh, test. A network error here is the URL; a 401 is the key. The curl test that separates them is in our network error and 401 guide.

What a message costs

The xAI model documentation, read on 28 September 2026, prices the current Grok models by request size:

ModelContextInput, per million (<200k)Output, per million (<200k)
grok-4.7500k$2.00$6.00
grok-4.6500k$2.00$6.00
grok-4.5500k$2.00$6.00
grok-4.31M$1.25$2.50
grok-build-0.1256k$1.00$2.00

Prices double once a request passes 200,000 tokens, which a Janitor request never approaches unless the chat memory is enormous. A 20,000-token message plus a 300-token reply on grok-4.7 or grok-4.6 is 20,000 at $2 per million, four cents, plus a fraction of a cent on output. On grok-4.3 the same message is about two and a half cents with far more context headroom. It sits between Claude, at roughly the same input cost, and Mistral or DeepSeek, which are pennies.

Cost of one 20,000-token Grok message with a 300-token reply, at xAI's listed prices
grok-4.7 / 4.6 / 4.5$2 in, $6 out per million
4.2 cents
grok-4.3$1.25 in, $2.50 out
2.6 cents
grok-build-0.1$1 in, $2 out
2.1 cents
Source: xAI model documentation prices for requests under 200k tokens, read 28 September 2026

The OpenRouter route

OpenRouter carried eight Grok ids on 28 September 2026. x-ai/grok-4.7 cost $1.60 and $4.80 per million, x-ai/grok-4.6 and x-ai/grok-4.5 $2 and $6, and x-ai/grok-4.3 $1.25 and $2.50. Proxy URL https://openrouter.ai/api/v1/chat/completions, your OpenRouter key, one of those ids, and a balance. None is :free. This is the route to pick if you already keep an OpenRouter balance for other models; the setup is in our OpenRouter guide.

The two caveats a proxy does not dissolve

Content. xAI's acceptable-use policy governs what Grok produces through the API, and it is stricter than the uncensored models Janitor users often choose. It also forbids accessing the service through unauthorised automated means, scraping or reselling outputs, and circumventing rate limits or safety measures. A refused scene is Grok declining under that policy. No route, and no IP address, changes it.

Region. The xAI terms require users to follow trade and economic sanctions laws. Its newer models have also launched EU-last: Grok 4.5 shipped on 8 July 2026 without EU availability, with xAI targeting mid-July. If the API does not serve your country, the terms decide, and we give no steps around them. If the model answers and then refuses a scene, that is the content policy, and no proxy is relevant.

Rate limits, and where a network proxy fits

xAI documents per-team rate limits in requests per second and tokens per minute, rising by tier, and returns 429 on excess. The limit follows the key, so neither a second key nor a different address moves it, the rule we set out for every backend in Janitor AI error 429.

A network proxy changes the address your traffic leaves from, and nothing on this page needs one. If xAI or Janitor AI does not serve your country, or your school or office network blocks it, we give no steps around that. Janitor adds one rule that follows the address. It asks users connecting from Australia, Brazil or the UK for a one-time age check, and "Detection is IP-based only." Where our lines do fit is a network that needs a known exit, such as a company network. For that, a static residential or ISP address in the country where you are costs from $0.44 per GB on a usage meter. Our proxies for Grok post covers the consumer side, including the national blocks Grok met in early 2026 and how they were lifted. For the price, the limits and the refusals on this page, a network proxy does nothing, and the honest setup is a console.x.ai key, the direct URL, and grok-4.7.

Sources

Frequently asked questions

What is the Grok proxy URL for Janitor AI?
https://api.x.ai/v1/chat/completions, with a key from console.x.ai and a model name such as grok-4.7. xAI documents its API as usable through the OpenAI SDK by setting the base URL to https://api.x.ai/v1, and its API reference lists POST /v1/chat/completions, which xAI now calls a legacy endpoint but still serves, so Grok connects to the Janitor proxy field with no relay. The OpenRouter route is https://openrouter.ai/api/v1/chat/completions with an id such as x-ai/grok-4.7.
Is Grok free on Janitor AI?
No. Grok's free tier lives in the X app and on grok.com, not in the API, and API access is paid per token. xAI listed grok-4.7 and grok-4.6 at $2 per million input tokens and $6 per million output for requests under 200,000 tokens on 28 September 2026, which makes a 20,000-token roleplay message about four cents. There is no :free Grok on OpenRouter. A free Grok proxy passed around on Discord runs on somebody else's key.
Which Grok model should I type into Janitor AI?
grok-4.7 is the newest model in the xAI model list on 28 September 2026, at $2 and $6 per million tokens under 200k, the same as grok-4.6 and grok-4.5; grok-4.3 is cheaper at $1.25 and $2.50 with a one-million-token context. On OpenRouter the ids are x-ai/grok-4.7, x-ai/grok-4.6, x-ai/grok-4.5 and x-ai/grok-4.3, and OpenRouter listed x-ai/grok-4.7 lower that day, at $1.60 and $4.80. Prices double above a 200,000-token request, so watch how much context Janitor sends.
Why might Grok refuse a scene on Janitor AI?
Because xAI's acceptable-use policy governs what Grok will produce through the API, and it is stricter than the models people usually pick for uncensored roleplay. A refusal is the model declining under that policy, not a broken proxy setting, and no route or IP address changes it. The policy also forbids scraping and circumventing safety measures, which matters if you were planning to automate around it.
Does Grok's API work everywhere?
Not necessarily. The xAI terms require users to follow trade and economic sanctions laws, and its newer models have launched without EU availability at first: Grok 4.5 shipped on 8 July 2026 without EU access, with xAI targeting mid-July for it. If the API does not serve your country, we give no steps around that; if the model itself refuses a scene, no address helps either.
Can a proxy raise Grok's API limits?
No. xAI documents per-team rate limits in requests per second and tokens per minute, set by how much you have spent, and returns 429 when you exceed them. The limit follows the API key, so a different IP address or a second key changes nothing. We give no steps around a limit or a block.

Get proxies that are alive right now

Our free proxy list re-checks every exit every few minutes across 100+ countries, with a live last-checked time, so you copy IPs that worked moments ago, not a stale text dump. When the location has to survive a real check, the paid network holds up.

129M+ proxy checks run · 100+ countries · HTTP / HTTPS / SOCKS · re-checked every few minutes · no signup

HProxy.

Honest guides and comparisons on proxies, scraping and staying unblocked, from the team that runs the network.

RSS feed