NewFree guided workflows for designers, PMs and developers
SkillHub
How to install

AI connectorsconnect the tools you use

What connectors (MCP servers) are, how to add one in Claude, ChatGPT, Gemini or a coding agent, and which to start with for your role. MCP is an open standard, so one connector works in all of them. No terminal needed for most.

01What connectors are

Accessto your files, tickets and data

A connector links your AI agent to another app, so it can read your Figma file, find the Linear ticket or open the pull request instead of you pasting it in. Most can also act there: create an issue, update a page, run a query.

Under the hood a connector is an MCP server: a small service that speaks the Model Context Protocol, an open standard. That's why one connector works in Claude, ChatGPT, Gemini and coding agents alike.

A connector gives the agent access; a skill tells it what to do with it. The Figma connector reads your file, a design-review skill knows what to look for.

02Add a connector

How to add a connectorwherever you work

Pick where you use AI. The connector is the same everywhere; only the menu differs. Each connector page on SkillHub links to its source with the server URL or command.

Use it in

In Claude

The Claude app on the web, desktop and mobile. No terminal, every plan.

Free plans get one custom connector. On Team and Enterprise, an owner adds connectors for the organisation first. Connectors you add here also show up in Claude Code.

  1. 01

    Open your connectors

    Customize → Connectors, then the + next to Connectors. Or, in a chat: + (bottom left) → Connectors → Manage connectors.

  2. 02

    Pick one and connect

    Browse the directory, open a connector and click Connect. Sign in to the app and approve what Claude may access. Not listed? Choose Add custom connector and paste the server's remote MCP URL.

  3. 03

    Turn it on in a chat

    + → Connectors, and switch on the ones this conversation needs. Then just ask: “summarise the open issues in our Q4 project”.

Tools on your own computer
The chat apps only reach servers on the internet. For local files or apps, use a coding agent, or in Claude Desktop the extensions: Settings → Desktop app → Extensions → Manage, then Browse extensions and install. No config file to edit.
Many connectors? Load on demand
Every active connector adds its tools to the chat, which costs context. Switch on only what a conversation needs. In Claude, with ten or more, set + → Connectors → Tool access to On demand.
03Where to start

One or two connectorsfor the tools you open every day

Start with the apps you already copy from most. Each one links to its page on SkillHub with the install details.

Browse all connectors

Designers

All connectors for designers

Product managers

All connectors for product managers

Developers

All connectors for developers
04Use them safely

A connector acts for youtreat it like a new colleague

Connect only what you trust
Connectors in an app's own directory are reviewed by that company; custom ones you add by URL are not. A connector can read and change data in the app it connects to, so check who built it first.
Give it the least access it needs
Review the permissions when you sign in and pick the narrowest scope. Turn off tools you don't use, especially ones that write or delete.
Approve actions until you know them
Your AI tool asks before it uses a tool or changes something. Choose “Allow always” only for tools you'd trust to run unsupervised.
Watch what comes back
Text from a web page, ticket or document can contain instructions aimed at your AI tool. If a tool starts doing something you didn't ask for, stop and check.
05Build your own

Your own serverfor developers

Have an internal API, database or tool without a connector? Wrap it in an MCP server. The official SDKs make a working server a few lines of code.

server.py
from mcp.server.fastmcp import FastMCP

mcp = FastMCP("my-server")

@mcp.tool()
def open_tickets(team: str) -> str:
    """List open support tickets for a team."""
    return fetch_tickets(team)  # your logic here

if __name__ == "__main__":
    mcp.run()
  1. 01

    Install the SDK

    pip install mcp for Python, or npm i @modelcontextprotocol/sdk for TypeScript.

  2. 02

    Describe each tool well

    The agent picks tools by their name and description, just like skills. Say what it does and when to use it.

  3. 03

    Test, then connect

    Try it in the MCP Inspector, then add it to your coding agent (claude mcp add, codex mcp add), or host it and add it by URL in any chat app.

Built something others can use? Submit it to SkillHub.

06Resources

Go deeperofficial docs and directories

Official documentation

Directories

Build your own

Find a connector for your tools

Every connector on SkillHub is reviewed, with what it connects to and how to add it.

Browse connectorsSkills guide