The MCP (Model Context Protocol) ecosystem now has over 14,000 servers. Most of them you do not need. Installing too many slows responses, inflates your context window with tool definitions that never get used, and turns debugging into a guessing game across a dozen integrations. This guide applies a simple filter: only install a server if it replaces a daily copy-paste workflow. It is organised by role — developers, QA testers, and DevOps/platform engineers — because the right stack is different for each.

Token usage with Claude Code follows a frustrating pattern: costs are not spread evenly — they cluster around a handful of bad habits. Most developers using Claude Code daily are burning 40–60% more tokens than they need to, simply because of how they phrase prompts, what they put in CLAUDE.md, and which model they reach for by default. This guide covers five concrete changes that make an immediate difference. Why Tokens Are Worth Caring About Every message you send in a Claude Code session includes:

The /loop command in Claude Code lets you schedule a prompt to repeat at a regular interval within an active session. It is one of the less-documented features, and it is also one of the most misunderstood — particularly around what it cannot do. This post explains what /loop actually is, the three scenarios where it genuinely earns its keep, and where cron jobs remain the better choice. What /loop Does /loop 5m check if the Vite dev server is still running and report any new errors This runs the prompt immediately, then repeats it every 5 minutes for the duration of your session.

Every CLAUDE.md file gets loaded into context on every session. Most teams treat it like documentation — a place to describe the project, list the tech stack, explain what the tests do. That is the wrong mental model and it is why most CLAUDE.md files are both too long and too ineffective. CLAUDE.md is behavioral programming. Its job is to change how Claude makes decisions, not to describe facts that Claude can read from the codebase itself.

There is a tendency in the AI tooling space to reach for frameworks — LangChain, AutoGen, CrewAI, OpenClaw — the moment you want an AI that does more than answer one question at a time. Most of the time, that is the wrong move. The framework adds complexity, dependencies, and debugging surface area for problems that a few shell scripts and cron jobs solve perfectly well. Claude Code’s headless mode (-p flag) plus a markdown file for personality plus cron scheduling is a complete AI assistant stack.