A working Claude connectors list: the official directory sorted by task, what the Verified, Community and Custom labels mean, step by step setup for individuals and teams, custom remote MCP servers, and the tool permissions that keep each connector in check.
You open a chat, ask Claude what is on your calendar tomorrow, and the answer comes from your real calendar instead of a guess. That is a connector at work. A connector links Claude to an outside app or service so Claude can find information there and take actions for you. This Claude connectors list sorts the options into two groups: the official directory you can browse today, and the custom connectors you can wire up yourself with a server URL. You will see what each category does, which label deserves your trust, how to add one in about a minute, and where the risks sit.
One note before the list. The directory ranks connectors by usage, like an app store, so the front page shifts from week to week. Treat the names below as a snapshot of what Anthropic's directory showed at the time of writing, and treat the setup steps as the part that stays stable.
Under the hood, a connector is a remote server that speaks the Model Context Protocol, usually shortened to MCP. Claude reaches it by URL, asks which tools it offers, and calls those tools when your request needs them: read a document, search a ticket queue, create an event. The tool list belongs to the developer who runs the server, not to Claude.
That is why the line between "official" and "custom" is thinner than it looks. In this article, official means listed in Anthropic's Connectors Directory, and custom means added by hand with a URL. Both kinds use the same technology and run on the same infrastructure. A directory listing makes a connector easier to find and gives it a review label. It does not change which tools the connector exposes.
Connectors work across the Claude apps, with small differences in what each app supports:
| Claude app | What it supports |
|---|---|
| claude.ai on the web | Remote connectors, including MCP Apps that show interactive content inside the conversation |
| Claude desktop app | Remote connectors, plus local connectors installed as desktop extensions |
| Claude mobile app | Remote connectors |
| Claude Code | Connectors you added in claude.ai, when you sign in with your Claude account |
| Cowork | Connectors and plugins |
💡 Tip: Add a connector once in claude.ai and it follows you into Claude Code, in the terminal and in cloud sessions, as long as you sign in with the same Claude account. There is no second setup.
The directory lives in two places: the public page at claude.com/connectors, and Customize > Connectors inside claude.ai and the desktop app. On the web page you can sort by Recommended, Most popular, Trending, Newest, or alphabetical order, and filter by category: Commerce & Shopping, Communication, Data & Analytics, Developer Tools, Financial Services, Healthcare, Legal, and more. Hundreds of connectors are listed, so searching by service name is usually faster than scrolling.
Here is the snapshot, grouped by what people actually ask Claude to do.
| Category | Connectors on the directory front page | A typical first request |
|---|---|---|
| Docs, mail, project work | Google Drive, Gmail, Google Calendar, Microsoft 365, Notion, Slack, Asana, monday.com, Box, Zoom | "What is on my calendar tomorrow afternoon?" |
| Developer tools | Atlassian, Linear, Supabase, Vercel | "Summarize the open issues assigned to me." |
| Design and creative | Canva, Figma, Adobe, Miro | "Find the latest brand assets for a launch post." |
| Sales, support, commerce | HubSpot, Intercom, Shopify, Indeed | "Which support conversations stayed open this week?" |
The web page also has Trending and New rows that rotate, so the table sticks to connectors from the top row of the page.
The Google trio (Drive, Gmail, Calendar) and Microsoft 365 sit at the top because they hold the material people ask about most: files, threads, and meetings. Notion, Box, and Slack extend that to team knowledge and conversations, while Asana and monday.com bring task boards into the same chat.
A good first test is a read-only question such as "What did I agree to in last week's email threads?" If the answer comes from your own mailbox, the connection works.
Atlassian, Linear, Supabase, and Vercel sit between writing code and shipping it: issues, databases, deployments. Developers get a second route too. Because Claude Code reads the connectors from your claude.ai account, the same Linear or Atlassian connection can answer a ticket question from the terminal without a separate install.
Canva, Figma, Adobe, and Miro bring visual files and boards into the conversation. These pair naturally with image and video generation, which the later section on PicassoIA walks through: Claude looks up the brief, and a separate image tool produces the visuals.
HubSpot, Intercom, and Shopify connect customer records, support conversations, and storefront data. Indeed adds hiring searches to the mix. These connectors touch customer information, so read the permission scopes on the sign-in page before you approve anything.
Every listing carries a label, and Anthropic's connector verification page spells out what each one means.
| Label | Who built it | Review level | Where you see it |
|---|---|---|---|
| Verified | Anthropic or a third-party provider | Anthropic tested the tools for quality and compatibility | A checkmark next to the name |
| Community | A third-party developer | Screened by Anthropic, not reviewed in depth | A Community label in the directory and in Customize > Connectors |
| Custom | You or your organization | Not reviewed by Anthropic | A Custom label in Customize > Connectors |
A Verified connector had its tools tested by Anthropic for quality and compatibility, and it met the Software Directory Policy requirements at the time of review. That is a real signal, but it is not a security audit. The developer still operates the connector and controls its tools, and those tools can change after the review.
A Community connector was built by a third party. Anthropic screens it before listing but has not reviewed it in depth, and Claude shows a reminder before you connect one. Once connected, it has the same capabilities and access as any connector you grant, so the label should shape how much you trust the developer, not how much access you give.
Developers who submit a server get an automatic policy scan, and by default the listing starts as Community. Listings flagged as highly useful to Claude users can be escalated to Verified review, where reviewers run a functional test of each tool. There is no separate application for the checkmark.
💡 Rule of thumb: Prefer Verified when two connectors do the same job. Use Community only for developers or companies you already trust.
You need a Claude account and an account with the service you are connecting. That service is run by its provider, and some providers require their own paid plan.
On these plans an Owner adds the connector for the organization, and each member then connects with their own account. Owners see Connect for your team. The arrow next to it offers Connect just for me, which lets you try a connector without adding it for everyone.
Members who lack permission see Request instead of a connect button, and the button changes to Requested while the request is pending. Owners see a Requested by your team section at the top of Organization settings > Connectors, plus a Requests tab under Organization settings > Notifications. From either place they can enable the connector or dismiss the request. On Enterprise plans, a member whose custom role includes managing the organization's libraries can add one too.
A connected connector is not automatically active in every conversation. In a new or existing chat, select + in the message box, choose Connectors, and use the toggle beside the one you want. When Claude is about to use a tool, it can ask for approval: Allow once continues this time, and Always allow skips the prompt for that tool from now on.
Turning the toggle off stops Claude from using the connector in that chat. The connector stays signed in, so you can turn it back on anywhere.
If the service is not in the directory and you have the URL of its remote MCP server, add it as a custom connector. It works on Free, Pro, Max, Team, and Enterprise plans, and the Free plan allows one custom connector. Custom connectors allow connections to unverified services, so the safety section below applies with extra force.
On Free, Pro, and Max:
https://mcp.example.com/mcp.On Team and Enterprise, an Owner goes to Organization settings > Connectors, selects Add, then Custom (choose Web if Claude asks for the type), and enters the URL. Members then find the connector with the Custom label and click Connect.
One network detail trips people up. Claude connects to your server from Anthropic's cloud, not from your laptop, so a server running on localhost will not be reachable. For servers inside a private network, Anthropic's documentation points to MCP tunnels.
| Authentication option | What happens | Pick it when |
|---|---|---|
| Sign in now | Each user signs in through the server's OAuth flow before using the connector | Every person needs their own account access |
| Sign in when needed | Claude connects without credentials and prompts for sign-in when the server asks | The server only needs credentials for some tools |
| No sign-in | Anyone with the server URL can use the connector | The server is public, or it relies on a fixed token in request headers |
For the OAuth client, the recommended choice is Use Claude's published identity, which needs no setup on your side but requires server support. Register automatically uses Dynamic Client Registration, which works with most servers. Use your own OAuth client takes a client ID you registered with the server.
Request headers are in beta for a limited set of organizations. They let you store a fixed credential, such as a shared token for an internal tool, and Claude sends it on every request. You can add up to four headers, and Anthropic approves each custom header name. Claude sends the value exactly as typed, so a bearer token needs the Bearer prefix included.
Two limits are worth knowing before you click Add. You cannot change authentication settings afterward, so a mistake means removing the connector and adding it again, and members must reconnect. And change the transport setting only if the server's documentation says to: a URL ending in /sse selects the older SSE transport.
Some tools should never leave your machine. For those, the Claude desktop app supports desktop extensions: local MCP servers packaged as .mcpb files that run on your device. Open Settings > Extensions, drag the file onto the page, review the preview, then select Install and confirm.
An extension runs with your user permissions, so it can reach only what you can reach. Your organization can turn extensions off or limit which ones you may install. Pick an extension for local files, sensitive data, or offline work, and a URL connector for internet-hosted services.
A remote server gives Claude tools that can read data, create or change it, delete it, and act on your behalf. Anthropic's own advice for connectors you add by URL boils down to a short list:
Open Customize > Connectors, select the connector under Your connectors, and find Tool permissions. You can set a level for a whole group of tools or for a single tool.
| Level | Behavior | Good fit |
|---|---|---|
| Always allow | Claude uses the tool without asking | Read-only tools on a trusted connector |
| Needs approval | Claude asks before each use | Tools that create, edit, or send |
| Blocked | Claude cannot use the tool | Tools you never need, such as delete actions |
This split is a habit, not a rule from the docs: let the reading happen freely, ask before anything changes, and block what you will never use.
/sse address to a /mcp address. Your connection keeps working, but moving to the new endpoint means removing the connector and adding it again from the directory. Adding it without removing the old one leaves you with two connections.
Connectors are not only for office tools. PicassoIA runs its own MCP connection, and the PicassoIA tools in claude.ai let Claude start image and video jobs straight from a chat. Four models sit behind it:
| Model | What it does |
|---|---|
| PicassoIA Image | Text to image |
| PicassoIA Image Editor Pro | Edits an existing image from a prompt |
| PicassoIA Video | Video from text or an image |
| Seedance 2.5 Lite | Video with audio |
Jobs are asynchronous. Claude starts a job, then checks its status until the image or video is ready. Each account runs up to five predictions at once, and that limit is shared across MCP connections, so a batch of ten prompts waits for free slots. Manage the connection from your PicassoIA account at picassoia.com/en/mcp/accounts, and check your plan on the pricing page to confirm MCP access. If the connector does not appear when you search the directory, use the custom connector route from the earlier section with the connection details from your account.
💡 Workflow idea: Use a design connector such as Canva or Figma to pull the brief, then ask Claude to generate three hero images with the PicassoIA tools and compare them in the same chat.
Connectors live inside the Claude apps. The Claude Sonnet 5 page on PicassoIA is a separate text workspace with a prompt, an optional image, an effort level, an output length, and a system prompt, and it has no connector setting. It is still handy for the writing around connectors: test plans, tool descriptions, and error triage.
Other Anthropic models sit in the same category, including Claude Fable 5, Claude Opus 4.7, and Claude 4.5 Haiku for fast drafts.
Every image in this article started as a plain sentence about a desk, a data closet, or a padlock on a gate. You can do the same in a few minutes. Pick a scene from your own work, such as a product on a counter or a team around a table, describe the light and the lens, and run it on PicassoIA Image. Refine the result with PicassoIA Image Editor Pro, then bring it to life with Seedance 2.5 Lite or PicassoIA Video.
Connect the tools you already use, keep permissions tight, and let Claude handle the errands. Then open Picasso IA and make something worth sharing.
