Connectors and Skills: connecting Claude to your team's work
In brief
Connectors let Claude reach the tools your team already uses. Skills teach it how your team does recurring work. What each one does, which to turn on first, and how to keep it manageable.
Contents
Out of the box, Claude is capable but isolated. It can write, reason, and analyse, but only with what you give it in the conversation. Two features change that, and they are easy to mix up:
- Connectors change what Claude can reach: your Drive, Slack, CRM, task tracker.
- Skills change what Claude knows how to do your way: your report format, your proposal template, your team's process for a recurring task.
Most teams need a little of both. Here is what each does and what to turn on first.
What Connectors actually do
A connector gives Claude read access — and sometimes action access — to an external service. Google Drive, Notion, Slack, Jira, Salesforce, Gmail, and many more. Instead of copying and pasting content into the chat, you ask Claude a question and it pulls the relevant information directly.
The practical difference: "Summarise the meeting notes from last Tuesday" just works, instead of requiring you to find the doc, open it, copy it, paste it, then ask.
Some connectors are read-only. Others let Claude take actions — creating tasks, drafting messages, updating records. The permission scope depends on what you grant when you set it up, and you can adjust individual permissions at any time in the connector's settings.
You don't need to leave all of them on
This is the thing nobody tells you: having every connector active in every conversation isn't necessarily better. When a connector is enabled, Claude considers it as a potential source for every query. That means more overhead, more token usage, and occasionally Claude pulling from a source that wasn't relevant to what you were asking.
A better approach: keep connectors active that apply to your current work context, and disable the ones that don't. Most connector interfaces let you toggle them without fully disconnecting — you're not losing the integration, just pausing it.
If you're doing writing work: Google Drive is relevant. Jira probably isn't.
If you're triaging bugs: Jira and GitHub are relevant. Your CRM probably isn't.
Switching contexts? Switch your connector set.
Connectors consume tokens
When Claude reads from a connector, the content it retrieves lands in your context window. A precise query that returns one relevant doc is fast and cheap. A vague query that returns fifteen documents — most of which aren't what you wanted — is slow and expensive.
This is why the way you phrase connector requests matters more than it might seem. "Find the Q3 roadmap in Notion" is a broad query. "Find the Notion page titled Q3 Product Roadmap in the Product workspace" is a specific one. The second costs fewer tokens, runs faster, and gives Claude something it can actually use.
For repeatable requests — daily task list, weekly pipeline summary, inbox triage — it's worth investing ten minutes to write a precise instruction once. Save it in your project instructions. Run it every time without rewriting it.
What's worth connecting first
Google Drive / Notion / Confluence — if your team stores docs there and you find yourself copy-pasting things into Claude, connect it. One setup, zero copy-paste from then on.
Jira / Linear / Asana — for anyone who needs to track or update tasks without switching tabs constantly.
Slack / Gmail — useful for context retrieval ("what did we decide about X?") but be deliberate about action permissions. Having Claude draft a reply is useful. Having it send without review is a different risk level.
Salesforce / HubSpot — high value for sales and CS teams who need account context without navigating CRM screens.
Start with one. The habit of working with a connected tool is what matters — not having everything connected at once.
What Skills actually do
A skill is a small package of instructions (and sometimes templates or scripts) that teaches Claude how your team does a specific thing. Claude only loads a skill when the task calls for it, so having many available costs nothing until one is needed.
Some come built in. Anthropic's document skills are why Claude can hand you a real PowerPoint deck, Excel file or Word document instead of a text outline.
The ones that matter most are the ones your team writes. Good candidates are anything you've explained to Claude three times: how you write a QBR summary, the format of your weekly update, the steps of your onboarding checklist. Turn that explanation into a skill once, and everyone gets the same result without pasting instructions.
How a manager sets them up:
- On Team or Enterprise, an admin can add a skill for the whole organisation, so it updates in one place when your process changes. Ask your admin before building a pile of personal ones.
- For a single team, put the instructions in your shared Project first. When the same instructions start being useful across several projects, that is the signal to turn them into a skill.
Our Skills setup guide covers the details, and Building skills for your team shows how to write a good one.
The practical starting point
- Connect the one tool where your team's documents live.
- Pick the one task you re-explain most often and put its instructions in your shared Project.
- Once that works, turn it into a skill so the rest of the company can use it too.
Further reading
- Introducing Agent Skills — what Skills are and how they work
- A complete guide to building skills for Claude — the full guide to creating and managing Skills
- Your favourite work tools are now interactive connectors inside Claude — how connectors bring external tools into Claude
- Discover tools that work with Claude — the connectors directory