Skills and Connectors: how to make Claude actually useful at work
By default, Claude only knows what you tell it in the conversation. Skills and Connectors change that — here's what they do and which ones are worth turning on.
Out of the box, Claude is a capable but isolated assistant. It can write, reason, and analyse — but only with what you paste in. It can't look things up, access your files, or take actions in other tools.
Skills and Connectors change that.
What's the difference?
Skills are built-in capabilities you enable inside Claude.ai — things Claude can do, not places it can reach. Web search. Code execution. Image generation. When you enable a skill, Claude gains access to that capability and will use it automatically when it's relevant to your conversation.
Connectors link Claude to external services — Google Drive, Dropbox, Notion, Jira, GitHub. When you connect a service, Claude can read documents from it, search it, or in some cases take actions within it. This is how you give Claude access to your actual work instead of making you copy-paste everything.
The practical distinction: Skills extend what Claude can do. Connectors extend what Claude can see.
Which ones are actually worth turning on?
Web search — almost always worth enabling. Claude's training data has a cutoff date. For anything time-sensitive — current events, recent product launches, prices, who currently holds a role — search is the difference between an accurate answer and a confidently wrong one.
Google Drive / Dropbox / Notion — worth it if your team stores documents there and you find yourself copy-pasting things into Claude regularly. Set it up once, stop doing that.
GitHub — valuable for technical teams who want Claude to work with their actual codebase rather than pasted snippets.
Code execution — useful if you want Claude to run calculations, manipulate data, or test code rather than just write it. Claude can write the code and verify it works in the same step.
What Connectors don't do
Connectors give Claude read access to your files and the ability to search within them. They don't give Claude persistent memory across conversations, and they don't let Claude proactively go looking for things without you asking.
Think of it like a smart intern who can access your Google Drive but only looks at things when you ask, and doesn't remember what they found in the last conversation.
For teams building more persistent AI workflows — where Claude should proactively pull context, take actions, or remember across sessions — that's a different setup, closer to what Claude Projects and the Agent SDK are for.
The practical starting point
If you're using Claude.ai for work: enable web search immediately. Then look at where your work documents live and connect that service. Those two things alone will make most conversations noticeably more useful.
Further reading
- Skills in Claude — Anthropic Support
- Available connectors — Anthropic Support