Claude Code persists project context with CLAUDE.md

// 87d agoTUTORIAL

Claude Code persists project context with CLAUDE.md

Claude Code loads CLAUDE.md files at session start so teams can persist coding standards, workflows, and architecture guidance across terminal and IDE runs. Anthropic’s memory docs position this alongside auto memory, giving developers both explicit rule files and learned preferences to reduce repeated prompting.

// ANALYSIS

This is a practical shift from prompt-by-prompt babysitting to repo-level operating instructions for coding agents.

–CLAUDE.md supports project, user, and org scopes, so teams can layer standards instead of restating them each session.
–The `.claude/rules/` and import system make memory modular, which is especially useful in larger monorepos.
–Auto memory captures recurring corrections, but explicit CLAUDE.md conventions are still the more controllable path for consistent outputs.
–Built-in commands like `/init` and `/memory` lower the setup cost and make this easy to adopt in existing workflows.

// TAGS

claude-codeai-codingagentcliprompt-engineering

DISCOVERED

87d ago

2026-03-02

PUBLISHED

87d ago

2026-03-02

RELEVANCE

9/ 10

AUTHOR

Better Stack

// KEEP READING

More AI developer news from the feed

EXPLORE FULL FEED

INFRA21m ago

Hippocratic AI hits 99.9% safety on NVIDIA Blackwell

Hippocratic AI achieved 99.9% clinical safety and a 2x prefill speedup using DigitalOcean’s NVIDIA Blackwell-powered AI-Native Cloud. The collaboration demonstrates the real-world performance gains of the HGX B300 for high-concurrency, safety-critical medical agents.

UPDATE25m ago

Claude Code adds automated fixes, persistent model defaults

Claude Code v2.1.153 introduces `/code-review --fix` to automatically apply suggested improvements and persists model selections as defaults. The update also ships critical security patches for OAuth credentials and resolves major memory leaks for long-running sessions.

NEWS45m ago

Midjourney founder: diffusion wins as FLOPS outpace memory

David Holz argues that diffusion models are the superior long-term architecture because they scale with cheap compute (FLOPS) while autoregressive models remain bottlenecked by expensive memory bandwidth.