# What is a context window in AI?

A context window is how much text, counted in tokens, a language model takes into account at once, instructions and documents included.

Updated on 22 October 2026.

## How it works

A model reads its whole context window to produce each answer. What sits outside it, such as a document nobody pasted or an earlier part of a long conversation that was dropped, does not exist for the model.

Larger windows hold more, but quality still depends on what is in them: the right three documents usually serve better than thirty loosely related ones.

## An example

Asked to update the onboarding guide, an assistant given the guide, the latest process change and the style rules does the job. Given the whole wiki, it may quote an outdated page.

## Common mistakes

- Pasting everything, which buries the documents that matter.
- Giving the model an old copy of a document that has since changed.
- Assuming the model remembers a previous conversation it cannot see.

## In Nodary

Agents connected to Nodary fetch only the documents a request needs, through search and reading tools, instead of a pasted copy. For a chat elsewhere, a context brief gathers up to 12 sources, shows its size, tells you which sources changed since you copied it, and goes anywhere as Markdown.

**What Nodary does not do:** it does not change a model’s context window, and a context brief is a snapshot you prepare, not a permanent memory.

- [Connect an AI agent](https://nodary.ai/docs/mcp)

Try it on your own documents. The Free plan needs no card. [Create a free workspace](https://nodary.ai/sign-up)

## Sources

- [Context windows, Claude documentation, Anthropic](https://platform.claude.com/docs/en/build-with-claude/context-windows)
- [Large language model, Wikipedia](https://en.wikipedia.org/wiki/Large_language_model)

## Related terms

- [AI agent](https://nodary.ai/glossary/ai-agent): An AI agent is a language model that pursues a goal in steps, calling tools, reading their results and deciding what to do next.
- [MCP (Model Context Protocol)](https://nodary.ai/glossary/model-context-protocol): The Model Context Protocol (MCP) is an open standard that connects AI applications to outside tools and data through one common interface.
- [Canon document](https://nodary.ai/glossary/canon-document): A canon document is the version a team has agreed is true: when documents disagree, it decides and the others defer to it.

https://nodary.ai/glossary/context-window