What is context engineering?

Updated Oct 26, 2026 · Sources listed at the end

Context engineering is the practice of choosing and maintaining the right set of information (tokens) that a model sees while it works: instructions, files, tool results and conversation history, not only the prompt you type. Anthropic calls it “the natural progression of prompt engineering”.¹

Why it matters

Anthropic describes context rot: as the number of tokens in the context window grows, the model’s ability to recall information from it decreases. That makes context a finite resource. Their guiding principle is to find “the smallest possible set of high-signal tokens” that makes the desired outcome likely.¹

Context engineering in coding tools

• Instruction files: CLAUDE.md, AGENTS.md and Cursor rules put stable project knowledge in front of the agent every session.²³
• Scoped rules: Cursor can apply a rule only when matching files are involved, keeping unrelated instructions out.³
• Tools instead of pasted data: MCP lets an agent fetch what it needs on demand.

Prompt engineering vs context engineering

Prompt engineering is writing and organising instructions. Context engineering covers everything else that lands in the window as well.¹ See prompt engineering for coding.

Tools mentioned

• Claude Code: paid from $20/mo
• Cursor: free plan, paid from $8/mo

Sources

1. Anthropic: Effective context engineering for AI agents (29 Sep 2025)
2. Claude Code docs: How Claude remembers your project (CLAUDE.md)
3. Cursor docs: Rules