WritingmateWritingmate Documentation

Message Limits

Writingmate measures text-model usage in counted messages. A short request uses one counted message. A request with a long chat history, large files, or a long response can use more than one.

Your allowance is shared by Writingmate chat and text requests made through the OpenAI-compatible API. It is tracked separately for Basic, Pro, and Ultimate model categories.

What counts as one message

One counted message covers up to 16,000 effective tokens. A token is a small piece of text that an AI model processes. In English, 16,000 tokens is roughly 12,000 words or 24 pages of plain text.

Writingmate counts:

  1. the tokens in your prompt, chat history, and attached context
  2. the tokens in the model's response
  3. cached prompt tokens at half their normal rate

Every request uses at least one counted message. Writingmate rounds requests above 16,000 effective tokens up to the next whole message.

Exact counting formula

cached_discount = floor(cached_prompt_tokens * 0.5)
effective_input = max(0, prompt_tokens - cached_discount)
counted_messages = max(1, ceil((effective_input + completion_tokens) / 16000))

For example:

  • A request with fewer than 16,000 effective tokens uses 1 counted message.
  • A request with 25,000 effective tokens uses 2 counted messages.
  • A request with 48,000 effective tokens uses 3 counted messages.

Example with cached tokens

Suppose a request reports:

  • prompt_tokens = 19000
  • completion_tokens = 3000
  • cached_prompt_tokens = 4000

Writingmate calculates:

cached_discount = floor(4000 * 0.5) = 2000
effective_input = 19000 - 2000 = 17000
effective_total = 17000 + 3000 = 20000
counted_messages = ceil(20000 / 16000) = 2
Diagram showing how Writingmate converts input, output, and cached tokens into counted messages

Writingmate chat and the OpenAI-compatible API use the same counting rule.

Current plan allowances

The current public plans include these daily allowances:

Model categoryBasic (free)ProUltimate
Basic models10 per dayUnlimited*Unlimited*
Pro modelsNot available50 per dayUnlimited*
Ultimate modelsNot available5 per day20 per day

An allowance applies to the whole model category, not to each model separately. For example, using two different Pro models draws from the same Pro-model allowance.

Unlimited* is subject to the Fair Use Policy. Large files, long chat histories, and Projects can make one request use several counted messages.

Legacy, promotional, and AppSumo plans can have different allowances. Open Profile Settings → Subscription, or use the model usage indicator in chat, to see the limits and remaining usage for your account.

When daily limits reset

Daily allowances reset at 00:00 UTC. The usage indicator shows a countdown and converts the reset time to your local time zone.

Starting a new chat does not reset your allowance. It can reduce future usage because the new request does not include the previous chat's history.

AppSumo plans use monthly message pools with daily pacing limits. The monthly pools reset on the anniversary of the plan's activation date.

Chat and OpenAI-compatible API usage

Text requests to these OpenAI-compatible endpoints use the same category allowance as Writingmate chat:

  • POST /chat/completions
  • POST /completions
  • POST /responses

They do not have a separate API message quota. Each request is assigned to the category of the model it uses.

If you connect your own OpenRouter API key, supported requests that use that key bypass Writingmate's standard model-usage limits. OpenRouter's billing and limits then apply.

Image and video usage

Image and video generation do not use counted text messages.

For current subscription plans:

  • Image generation is controlled by plan access. It does not reduce the Basic, Pro, or Ultimate text-message allowance.
  • Video generation uses the plan's monthly video allowance. Queued, in-progress, and completed videos count toward that allowance; failed videos do not.

For AppSumo plans:

  • A successful image generation uses one image credit. A failed generation is refunded.
  • A video request reserves the requested number of seconds from the video-seconds pool. The reservation is refunded if the request fails to start.
  • Image and video pools reset on the anniversary of the plan's activation date.

The same media rules apply when you use the OpenAI-compatible POST /images/generations and POST /videos endpoints.

Use messages efficiently

  • Start a new chat when earlier history is no longer useful.
  • Attach only the files and sections needed for the task.
  • Split large Projects so unrelated files are not sent with every request.
  • Use a Basic model for work that does not need a Pro or Ultimate model.
  • Check the model usage indicator in chat before a large file or API workflow.

Related docs