Practical. No hype.
The most common mistake I see surgeons make with AI is using Chat for everything. Chat is fine. It’s where most people start and it’s a genuinely useful tool. But it’s one mode of one product, and using it for tasks that other modes handle better is like doing everything with a scalpel because it’s the instrument you picked up first.
Claude has several distinct modes. Understanding what each one is for takes about ten minutes and makes the difference between using AI occasionally and using it well.
Chat
Claude Chat is the conversational interface — the tab you open on your browser, the app on your phone. It’s where you have back-and-forth conversations, ask questions, get drafts, work through ideas. Within a Project (read the full setup guide here), it has persistent context. Outside a Project, each conversation is standalone.
Best for: drafting, summarising, answering questions, reasoning through a problem, generating ideas. Most clinical and administrative tasks live here.
Cowork
Claude Cowork is the desktop application. It does what Chat does, plus more: it can access files on your computer, run code locally, interact with documents, and execute multi-step tasks without you staying in the loop at each step. It’s the agentic version — Claude working on your behalf rather than just answering your questions.
In practice for surgeons: Cowork is where you’d ask Claude to take an Excel spreadsheet of your rota, parse the shift dates and times, and automatically add them to your calendar — the kind of task that involves multiple steps across different tools. Chat can draft an email. Cowork can draft the email, find the recipient’s address in your contacts, and queue it for sending.
Best for: multi-step tasks, file manipulation, anything that involves more than one action in sequence. Included in Pro and above.
Claude Code
Claude Code is a command-line tool for software development — it runs in a terminal and can write, edit, test, and debug code directly in your file system. For most surgeons, this is not relevant. It becomes relevant if you’re building something: a data analysis pipeline for audit, a spreadsheet automation, a simple web tool. If you ever find yourself thinking “I wish I could automate this,” Claude Code is how you build that automation without being a programmer.
Best for: building tools, writing code, technical projects. Not a day-to-day clinical tool.
Skills
Skills are pre-built workflows that run inside Cowork. Rather than starting from scratch and instructing Claude step by step, a Skill handles a specific recurring task — “process my inbox,” “convert this rota to calendar events,” “triage this document” — with all the steps already defined. You trigger the Skill and it runs.
The calendar skill I use is an example: I pass it a rota spreadsheet and it creates all the calendar events automatically. The Skills library has a growing number of pre-built options; you can also create your own for tasks you do repeatedly.
Best for: recurring structured tasks where you want a consistent process rather than a fresh conversation each time.
Plugins
Plugins connect Claude to external services — your calendar, your email, your documents. Once connected, Claude can read and act on those services as part of a conversation or a Skill. A calendar plugin lets Claude check your availability and create events. An email plugin lets it search your inbox or draft and send messages.
Best for: connecting Claude’s capabilities to the services you already use. Most useful in Cowork.
Prompts
A prompt is simply the instruction you give Claude. Everything above depends on good prompting — being specific about what you want, what format you need it in, what to avoid. The most common prompting mistake is asking a vague question and hoping for a specific answer. “Write me a letter about this patient” will produce a generic letter. “Write a referral letter to the pain team for a 58-year-old with chronic shoulder pain post-arthroplasty, two years out, failed conservative measures — clinical question is whether nerve block would help prior to revision discussion” will produce something useful.
For a practical guide to prompting, Anthropic’s prompting guide is the most reliable reference.
The decision guide
Quick question → Chat. Structured writing task in your voice → Chat in a Project. Multi-step task involving files or tools → Cowork. Recurring task you want automated → Skill. Technical build → Claude Code. Connecting to another service → Plugin.
That covers ninety percent of what a surgeon will actually use. Start with Chat and a Project. Add Cowork when you find yourself wanting Claude to do more than one thing in a row.
Next: Building your AI tribe — who to add, what to give them, how to start.
Not sure which mode to use for your specific situation? Logged-in members can book a session — 30 minutes (£75) to map your workflow to the right tools, or 60 minutes (£150) to set up the full stack properly so each mode is doing what it’s best at.
1 thought on “Chat, Cowork, Code, Skills, Plugins, and Prompts: what is what and when to use each”