Skip to content

Claude is only as good as the context you give it. Here’s how to structure that context.

Claude doesn’t know you. It doesn’t know your project, your team, your constraints. Every conversation starts from scratch (unless memory is enabled). It’s up to you to give it the framework it needs to work within.

Good context answers:

  • Who — your role, the people involved
  • What — the subject, the goal
  • Why — the motivation, what’s at stake
  • How — the expected format, the constraints

Sufficient for simple, non-critical tasks.

Fix the typos in this text: [text]

For most professional tasks.

I'm a project manager at a tech consultancy. I need to send a meeting recap
to a banking client following a production incident.
Write an email: professional, factual, reassuring. Max 200 words.

For complex or recurring tasks.

Context: I'm working on a Node.js/Express API in production.
Stack: Node 20, TypeScript, PostgreSQL, hosted on AWS Lambda.
Team: 3 devs, code always reviewed.
Conventions: strict ESLint, no TypeScript any, Jest tests required.
Task: [your specific request]

For projects you work on regularly, create a context block you paste at the start of each conversation:

## My work context
- Role: [your role]
- Project: [short description]
- Stack / tools: [technologies]
- Constraints: [non-negotiables]
- Expected tone: [how you want it to respond]

Next step → Anatomy of a good prompt