The context window is a language model's working memory: everything it sees while preparing an answer. That includes the prompt, the conversation history, documents and tool results. Its size is measured in tokens.
Once the window fills up, the model loses details from earlier in the conversation or can't take the request in full. That's why it pays to keep an agent's context free of clutter.
How Pachca accounts for this:
- Each thread sticks to one topic, so the agent doesn't mix up discussions
- Pachca CLI shows commands via --help, so docs don't eat up the context window
- Agent Skills are split by area of the API, so the agent loads only what it needs
- Every page on dev.pachca.com is available in Markdown: just add .md to the URL
For developers
Docs on dev.pachca.com →