Projects & Context

Projects give Claude persistent memory and custom instructions for your entire workflow.

What are Claude Projects?

Projects are the killer feature of Claude Pro and Team. A Project is a named workspace where you can:

Write a system prompt that applies to every conversation in the project
Upload shared files (code, docs, data) that Claude has in context for every conversation
Have multiple conversations that all share the same knowledge base
Collaborate with team members (Team plan)

Creating a Project

In the left sidebar, click "New Project". Give it a name (e.g., "Python codebase review", "SQL learning", "Data model design"). You'll see two configuration areas:

Project Instructions — your system prompt (more on this below)
Project Knowledge — files you upload that Claude references throughout the project

Writing an effective system prompt

The system prompt is the most powerful tool in Claude Projects. It runs before every conversation in the project, establishing who Claude is, what it knows, and how it should respond.

SYSTEM PROMPT TEMPLATE

You are a senior data engineer reviewing code and data models for [COMPANY NAME].

Context:
- Our main data warehouse is PostgreSQL 15 on AWS RDS
- We use dbt for transformations, Airflow for orchestration
- Naming convention: snake_case, singular table names, customer_id not id
- We normalize to 3NF in OLTP, star schema in the warehouse

When reviewing SQL or data models:
1. Flag any violation of our naming conventions
2. Check indexes on all FK columns
3. Suggest performance improvements
4. Note any data type mismatches (especially FLOAT for money — we use NUMERIC)

Respond concisely. Use code blocks for all SQL examples.

What to upload to Project Knowledge

Files in Project Knowledge persist across all conversations. Good candidates:

Schema files — your database DDL or data model diagrams as text
Style guides — coding standards, naming conventions, architecture docs
Reference docs — API documentation, configuration files, onboarding guides
Sample data — anonymized CSV files for analysis tasks
Codebase overview — a text file describing your codebase structure

WHAT NOT TO UPLOAD Never upload real PII (customer names, SSNs, etc.), production credentials, API keys, or confidential business data. Claude's API calls are processed on Anthropic's servers — treat uploads like you would any cloud service.

Project use cases for data professionals

SQL review project — upload your schema, write a system prompt about your conventions, ask Claude to review every new query you write
Data model design — upload requirements docs, ask Claude to help design entities and relationships
Python automation — upload your existing scripts, ask Claude to refactor, test, or extend them with awareness of your patterns
Interview prep — create a project with your resume and target JDs; ask Claude to run mock interviews
Blog writing — upload your writing style guide; Claude maintains consistent voice across posts

← The Claude Interface Artifacts & Code →