Glossary of Terms

This glossary defines the terms used in the current Writingmate app and documentation.

A

Agent: A saved AI assistant for a repeated task. An Agent combines one selected model with saved instructions and can also include knowledge files, tools, integrations, and conversation starters. A saved Agent is different from Agent Mode. See Overview of Agents.

Agent Mode: A chat mode for multi-step research, analysis, and planning. Writingmate can plan the task and use tools while working through one conversation. Agent Mode is not a reusable Agent card.

AI model: A system that processes a prompt and produces an output, such as text, an image, or a video. Writingmate provides models from multiple AI providers.

API (Application Programming Interface): A defined way for one application to send requests to another application. Writingmate provides an OpenAI-compatible API for supported external clients.

API key: A secret token used to authenticate an API request. In Writingmate, a provider API key and a Writingmate Developer Key have different purposes and are not interchangeable.

Assistant (legacy term): An older name for what the current Writingmate interface calls an Agent. Current documentation uses Agent.

B

Basic, Pro, and Ultimate model categories: The usage categories assigned to text models. Each category has its own access rules and shared message allowance for the current workspace plan. A model category is not the same as a subscription plan. See Supported AI Models.

BYOK (Bring Your Own Key): Using an API key from a provider account that you manage and pay for separately. Writingmate currently supports an OpenRouter API key for BYOK. Supported requests use that OpenRouter account, while Writingmate plan locks and other feature requirements can still apply. See Set Up Your API Keys.

C

Chat: A saved conversation between you and one or more AI models. A chat contains your messages, model responses, attachments, and enabled tools.

Chat Mode: The standard mode that sends your prompt to one selected text model. Use it for normal questions, writing, code, and analysis.

Compare Mode: A mode that sends the same prompt to two to four selected text models and displays their responses side by side. See Chat Overview.

Composer: The message box labeled Send a message... and its controls for adding content or tools, changing mode, choosing a model, using voice input, and sending a prompt.

Context: The information available to a model for the current request. It can include your prompt, chat history, Agent instructions, attached files, tool results, and profile memory.

Context window: The maximum amount of context a model can process in one request. The limit is measured in tokens and differs by model.

Counted message: Writingmate's unit for text-model plan usage. One counted message covers up to 16,000 effective input and output tokens. Every request uses at least one counted message, and a large request can use more than one. See Message Limits.

D

Developer Key: A password-like token that lets an external client use Writingmate's OpenAI-compatible API or MCP endpoint with your Writingmate account. Developer Keys begin with wm_v2. and use the current workspace's access and allowances. They are different from an OpenRouter BYOK key. See OpenAI-Compatible API.

F

File: A document, spreadsheet, text file, or image uploaded to Writingmate. Supported documents can be processed into text for a model. Image attachments require a model that accepts image input.

I

Images: The Writingmate workspace for creating, editing, saving, and downloading generated images. Images is also a tab in the model selector. Available models and controls can change. See AI Image Generation.

Integration: A connection that lets Writingmate or an Agent use an outside service or built-in tool. MCP servers are one type of integration.

K

Knowledge base: Saved reference files that can be reused in chat or attached to a saved Agent. In chat, select Add from knowledge base to choose a processed document from Knowledge Management. See Using Files.

M

MCP (Model Context Protocol): A standard that lets an AI application connect to outside tools and data sources. In Writingmate, you configure an MCP server, Writingmate discovers its tools, and a compatible model can call enabled tools. See MCP Integration.

MCP server: A service that exposes tools through MCP. A server can require a URL, headers, an API token, or OAuth authentication.

Message: A prompt you send or a response a model returns in a chat. A visible message is not always the same as a counted message used for plan limits.

Model selector: The control for choosing available text, image, or video models. In Compare Mode, it lets you select two to four text models. Agent Mode can hide this control while Writingmate chooses models for the task.

O

OpenAI-compatible API: Writingmate's API surface for clients that support OpenAI-style endpoints and Bearer-token authentication. Its base URL is https://writingmate.ai/api/openai/v1. It uses a Writingmate Developer Key. See OpenAI-Compatible API.

P

Plugin (legacy product term): An older product label for features now presented as Tools or Integrations. Some documentation navigation and URLs still use Plugins as a category name. Current app labels include Web search, Integrations, and MCP. See Tools and Integrations Overview.

Project: A workspace container for related chats and reference files. Projects appear only when the feature is enabled for the workspace.

Prompt: The instruction or other input you send to an AI model. A prompt can include text, attachments, and tool instructions.

S

Skill (saved prompt): A reusable instruction saved in Writingmate. The sidebar currently labels this collection Prompts, while Manage Data uses All Skills for the same saved records.

T

Token: A small piece of text that an AI model processes. Tokens can represent a whole word, part of a word, punctuation, or formatting. Writingmate uses input and output token counts when it calculates counted messages.

Tool: A capability a compatible model can call while answering, such as Web search or a function exposed by an MCP server. Enabling a tool does not guarantee that every model can use it.

V

Video: The Writingmate workspace for creating short videos from a text prompt or, when supported, a reference image. Video models have different duration, resolution, aspect-ratio, and audio options. See AI Video Generation.

Voice input: The composer control that records speech and converts it into text for your prompt. It does not send the message until you review and submit it.

W

Web search: A tool that lets a compatible model retrieve current information from the web while answering. See Web Search.

Workspace: The account context that holds a plan, members, usage, settings, chats, and shared resources. Billing and plan limits apply to the currently selected workspace.