API and limits

An OpenAI-compatible API: the address, the models, the limits your tools work within, and what every error means.

Connect

Set three things in any tool or library that speaks the OpenAI API: the base URL, your key (from the console) and a model.

Base URL: https://api.thecontext.club/v1
Models: qwen3.8-27b thinks before it answersqwen3.8-27b-fast answers straight away

Keep one model for a whole conversation: switching between them reads the conversation afresh. Tools that send prompt_cache_key keep each conversation's cache apart.

Base URLhttps://api.thecontext.club/v1API keycc_… (your key)Modelqwen3.8-27b

Works in Cline, Continue, Aider, OpenCode and any tool that takes an OpenAI-compatible URL.

Limits

By plan. Priority tokens reset at midnight NZ; cached prompt tokens count at a tenth.

	Community	Pro	Pro Max
Priority tokens a day	5M	10M	20M
Requests at once	2	4	5
Requests waiting for a place	8	8	8
Requests a minute	120	120	120

Every model takes prompts of up to 262,144 tokens and replies of up to 32,768 tokens. A priority request waits up to 55 seconds for a place; a standard request up to 10 minutes.

Choosing a queue

Requests use your priority tokens until they run out, then carry on in the standard queue. To send work to the standard queue yourself, add service_tier="flex" to a request, or use a key marked "Standard queue only" in the console. See How it works.

Response headers

`x-club-queue`	priority or standard: the queue that served this request.
`x-ratelimit-limit-tokens`	Your plan's priority tokens a day.
`x-ratelimit-remaining-tokens`	Priority tokens left today. At 0, requests carry on in the standard queue.
`x-ratelimit-reset-tokens`	Seconds until your priority tokens reset, at midnight NZ.
`x-ratelimit-limit-requests`	Requests a minute your plan allows.
`x-ratelimit-remaining-requests`	Requests left in the current minute.
`Retry-After`	On a 429 or 503: how many seconds to wait before trying again.

Errors

Errors come in OpenAI's shape, with a code. Give your tool a 15-minute timeout and plenty of retries, for example OpenAI(timeout=900, max_retries=20).

Status	Code	What to do
400	`model_not_found`	The model name is wrong. GET /v1/models lists the names.
400	`invalid_service_tier`	service_tier must be "flex" or "standard" for the standard queue, or "auto", "default" or "priority".
400	`context_length_exceeded`	The prompt is longer than the model's context. Shorten it.
401	`invalid_api_key`	The key isn't valid or was revoked. Create a new one in the console.
402	`subscription_inactive`	Your plan isn't active. Choose a plan or update your card under Billing.
413	`request_too_large`	The request is over 8 MB. Send less at once.
429	`rate_limit_exceeded`	Too many requests in the last minute. Wait for the Retry-After, then try again.
429	`queue_full`	You already have the most requests waiting that your plan allows. Let some start first.
429	`queue_timeout`	The club is busy and no place came free within a minute. Try again in a few seconds.
429	`standard_queue_busy`	A standard request waited 10 minutes without a place. Try again in a couple of minutes.
503	`launching_soon`	The club hasn't opened yet. Your key works from opening day.
503	`node_parked`	The GPU is switched off for the night and standard requests can't switch it on. A priority request will, or try after 7am NZ.
503	`node_waking`	The GPU is starting, which can take a few minutes at night. Try again in a few minutes.
503	`standard_resting`	The standard queue is paused until 7am NZ. Priority requests still run.
503	`spend_cap`	The club has paused for the rest of today. It is back at midnight NZ.
503	`node_paused`	The GPU is paused for maintenance. Try again later.
503	`node_unavailable`	The GPU isn't ready to take requests. Try again in a minute.
503	`store_unavailable`	The club's database is unavailable. Try again shortly.
502	`node_error`	The model stopped answering, or a reply was cut off partway. Send the request again.