Claude Code Tips, Setup & Productivity (68)

session-archive is a stdlib-only Python CLI that gzips, indexes, searches and restores Claude Code session logs before the default 30-day auto-delete.

HN discussion of Anthropic's blog post on how the Claude Code team handles code review.

A small offline test running the same local models on a 6GB laptop GPU through pi, Codex and Claude Code, where pi and Codex passed 12/12 tasks and Claude Code 7/12.

How to give Claude Code a second repo as read-only reference with --add-dir plus an Edit deny rule, since added directories follow the same edit permissions as the main one by default.

Why CLAUDE.md/AGENTS.md files fail as team memory (bloat, no recall, merge conflicts) and six practical fixes: rules-only files, decision logs, investigation notes, PR review, pruning, and shared search.

Part 1 of a Portuguese-language Claude Code CLI guide built from the official docs, covering install, entry commands, shortcuts, slash commands, permission modes, CLAUDE.md and settings.json.

Four common ways Claude Code context gets polluted, with fixes using /clear, /rewind, subagents and a lean CLAUDE.md.

How to rename an API field across three separate repos in one Claude Code task using --add-dir, a repo map and a stale-caller grep.

A Show HN tool that replaces Claude Code's regex code search to cut agent token use roughly in half.

How the author shares context between Claude Code and Codex on the same repo without the two agents overwriting each other.

A team logged 13 full runs of its autonomous Claude Code agent and found a median of 79.8M tokens per run, about 97% of them cache reads and 0.2% output.

Setting up the official Claude Code VS Code extension: install, permission modes, checkpoints and recommended day-one settings.

The author's Claude Code setup: one rulebook file plus a 12-step daily routine for using it consistently.

Walks through Claude Code Router v3.1, now a desktop app and local gateway, covering providers, model tiers, routing scripts and fallbacks.

When to keep work in the main Claude Code conversation versus moving it to a subagent, second session or headless loop.

Anthropic's guide to getting the best results from Opus 5.5 in Claude and Claude Code.

Why output quality drops as Claude Code's context fills, and how to manage it with /compact, /clear, /rewind and subagents.

A tip to back up your Claude Code configuration (CLAUDE.md, skills, agents, settings) since it grows into real infrastructure.

Five specific ways Claude Code struggled on a 240,000-line WordPress plugin codebase, observed during a course.

A test of whether a one-line fact in Claude Code's auto memory steers the next session as reliably as CLAUDE.md does.

A case study of adding OpenTelemetry tracing to 47 services in nine days with Claude Code.

How to shape Claude Code's output format so replies stay readable and concise.

Why Claude Code writes GNU-style shell commands that fail on macOS, and how to make it target BSD tools.

A case study migrating 90 Cypress tests to Playwright in four days with Claude Code.

What Claude Code's permission modes, --allowedTools, allowlists and sandbox boundaries actually restrict.

Spotify engineering's write-up on how its Portal tool cut Claude Code token usage by about 90%.

A comparison of Claude Code's built-in memory with the author's own memory system and the different jobs each does.

A case study cutting a 2.1 MB JavaScript bundle to 890 KB with Claude Code.

Why Claude Code reported a saved PDF as empty, and how to get reliable text extraction.

A Claude Code session that got macOS printing natively to an HP Laser 1008a printer.

Anthropic's guidance on getting more value out of each Claude Code session.

A decision guide for when to use Claude Code versus the Claude Agent SDK.

What Claude Code's auto permission mode approves now that it is the default, and how to switch back.

Anthropic's announcement that auto mode is now the default permission mode in Claude Code.

Claude Code documentation for sending messages between your running Claude Code sessions.

A case study using Claude Code to track down a production memory leak.

What headless claude -p loads on each run and when the --bare flag saves tokens.

A discussion of whether Opus 5 makes a large CLAUDE.md unnecessary, prompted by an interview with Claude Code's creator.

A setup guide for running Claude Code through OpenRouter.

A step-by-step guide to setting up a spare Mac for Claude Code to control.

How a question from a remote, long-running Claude Code task went unseen for hours, and how to avoid that trap.

Measurements showing raw web pages cost Claude Code tens of thousands of tokens, and how to fetch research more cheaply.

Workflow patterns from running a live automated trading system on Claude Code.

Using Claude Code to add auto-reconnect and other refinements to a macOS meeting translation app.

Common ways developers waste tokens with Claude, and habits that cut the cost.

Why a headless Claude Code agent can get stuck on 401 errors with a static OAuth token, and how to fix it.

Where Claude Code and Codex log token usage locally and how to read those files.

What one day of experimenting with Claude Code on AWS Bedrock cost, and where the money went.

A division of labor for running Claude Code and Codex side by side on the same work.

What actually proved useful after three weeks of running Claude Code on a real production project.

Pointing Claude Code at a custom backend with ANTHROPIC_BASE_URL, plus a checklist of what can break.

One addition to a Claude Code workflow that cut the cost of the same task from $1.96 to $0.74.

2026 migration guide from GitHub Copilot to Claude Code: pricing comparison, install, workflow mapping, MCP servers for extensions, and CLAUDE.md setup.

How long Claude Code projects accumulate "agentic technical debt" and ways to manage it.

Claude Code configuration options found in the source that the docs don't cover.

A guide to using Claude Code daily with CLAUDE.md, skills, subagents, plugins and MCPs.

Moving Claude Code from a laptop onto shared cloud compute a whole team can use.

Building and deploying an Astro blog with Tailwind, MDX, RSS and a sitemap from one Claude Code prompt.

Why Claude Code-built sites look alike, and how a custom image layer makes them distinct.

Anthropic's best practices for using Claude Code in large codebases.

Building a free S&P 500 technical-analysis app in a weekend with Claude Code as a pair programmer.

The case for having Claude Code produce HTML output instead of Markdown, with examples.

How to structure agent memory beyond CLAUDE.md so it survives context-window limits.

How an overflowing context window broke a long-running Claude Code agent, and how to spot it.

Seven rules that keep Claude Code from drifting and make it more useful.

A Claude Code debugging workflow for diagnosing and fixing production issues faster.

Testing strategies for replacing a manual QA routine with Claude Code-driven tests.

Five harness layers beyond CLAUDE.md (memory, tools, permissions and more) for production Claude Code.