Перейти к содержанию
TL;DR
Claude connectors let Claude read from and write to the tools you already use, without you pasting anything in. There are hundreds in the directory now, so rather than list them all, this covers twelve that cover the most common workflows – documents, meetings, project management, CRM, design, and finance – and explains what each one gives you.
| # | Connector | What Claude can access or do | Лучшее для |
|---|---|---|---|
| 1 | Диск Google | Docs, Sheets, Slides, PDFs and file search | Files and workspace knowledge |
| 2 | tl;dv | Meeting transcripts, notes and cross-meeting search | Сбор разведывательной информации |
| 3 | Понятие | Pages, databases and workspace content | Team knowledge and documentation |
| 4 | Slack | Messages, channels, threads, files and messaging tools | Team communication and decision context |
| 5 | Linear | Issues, projects and product workflows | Product and engineering workflows |
| 6 | GitHub MCP | Repositories, code, commits, issues and pull requests | Software development workflows |
| 7 | HubSpot | Contacts, deals, companies and tickets | CRM and revenue insights |
| 8 | Асана | Tasks, projects and goals | Project and task management |
| 9 | Figma | Code-to-design, design-to-code and design systems | Design and development workflows |
| 10 | Canva | Search, create, autofill, resize and export designs | Content and design creation |
| 11 | Полоса | Customers, payment intents and Stripe API tools | Payments and Stripe operations |
| 12 | Zapier | Search and actions across thousands of apps | Cross-app automation |
Оглавление
How I Picked These Connectors
I reviewed these connectors by usefulness rather than popularity. The question was not just how much a connector can do, but whether connecting it gives Claude useful context or actions for work people actually do.
Three things mattered most:
- What Claude can actually do with it: I looked at whether the connector is read-only, read and write, or Interactive, and whether those capabilities translate into useful workflows.
- Where it works: remote connectors in Claude’s directory work across Claude.ai, Desktop, Mobile, Claude Code and Cowork. Local MCP integrations are limited to Claude Desktop and Claude Code.
- What it costs to use: I also considered whether the connector works on free plans or requires a paid Claude or third-party subscription. A useful connector is harder to recommend broadly if you need two upgrades before you can use it.
I verified the details against Claude’s current Connectors Directory, Anthropic’s documentation and the connector providers’ own documentation.
For a wider look beyond Claude’s directory, tl;dv also keeps a running list of the best MCP servers.
What Is a Claude Connector?
A Claude connector is a hosted integration that lets Claude call another service’s tools during a conversation. Every one of them runs on the Model Context Protocol, the open standard Anthropic published and other AI clients have since adopted.
Claude inherits your permissions from the connected service. If you cannot see a file, channel, or record in the source system, the connector cannot reach it either.
Two routes exist. Directory connectors are added in one click from Claude’s own catalogue. Custom connectors are any remote MCP server you add by URL, covered step by step in how to add a custom connector in Claude. Both run on the same infrastructure once connected.
12 Best Claude Connectors (Reviewed)
1. Google Drive: Best for Documents and File Search
Google Drive is still one of the first connectors I’d add because it gives Claude direct access to files you already work with. Claude can search across Drive, read Docs, Sheets, Slides, PDFs, images and Microsoft Office files, retrieve file metadata and permissions, upload files, create folders, and save Claude-generated files back to Drive. Sheets are provided to Claude as CSV data, while Slides are extracted as plain text.
Drawback: Claude reads Google Sheets as CSV and Google Slides as plain text, so it does not get the full native file experience. It also extracts text only from Drive files, which means embedded images are not processed, and the connector does not support adding comments.
2. tl;dv: Best for Meeting Transcripts and Cross-Meeting Search
tl;dv gives Claude access to meeting context that usually never makes it into a document. Claude can pull transcripts and notes from Google Meet, Zoom and Microsoft Teams, summarize individual calls, and search across multiple meetings for a topic or decision.
I’d use it mainly for questions that span several calls. For example, you could ask which objections appeared repeatedly in demos last month, find every meeting where a pricing change was discussed, or compare decisions from product meetings with engineering stand-ups. By default, tl;dv searches meetings you participated in, but you can explicitly ask Claude to include other meetings you can access. tl;dv recommends using clear timeframes, and meeting searches return 50 meetings at a time.
Both routes are read-only, so Claude can pull a transcript but cannot start a recording or change anything. The directory connector’s consent screen grants access to meeting metadata, transcripts, and notes, which I confirmed when connecting it in August 2026. It needs tl;dv’s Pro tier or above, so a free account sees an upgrade prompt instead of the connect flow. The free plan has no cap on recorded meetings, though, so you can build a library first and connect later.
For a closer look at how the integration works and the kinds of cross-meeting queries it supports, the tl;dv MCP goes into more detail.
Drawback: The connector is read-only, and tl;dv Pro or higher is required. Access to individual meetings also depends on the meeting organizer having an eligible paid tl;dv plan. A paid Claude plan is not currently required for remote MCP connectors.
I’ll be honest about the competition, since this is the category tl;dv sells into. Both Otter and Fireflies has Claude connectors too, and Fireflies was the first meeting tool in Claude’s directory. If your team already runs on one of them, connecting what you have beats switching notetakers for a connector. The Fireflies review and the Otter.ai review cover where each lands. Tactiq is a different case: it connects through a custom connector URL, and Claude can search and list your Tactiq meetings on any plan but needs Tactiq’s Team plan to read full summaries. The Tactiq alternatives comparison goes further on this.
3. Notion: Best for Team Knowledge and Documentation
Notion works well when company knowledge lives across internal wikis, project pages and databases. Claude can search the workspace, retrieve information, create and edit pages, update page properties and organize content directly from a conversation. The connector is currently read and write.
If most of a team’s documentation already lives in Notion, I’d rather connect that existing knowledge than keep moving information manually into Claude.
Drawback: Notion search is currently limited to 30 requests per minute, while overall MCP usage averages up to 180 requests per minute across tools. That is unlikely to affect ordinary use, but it can matter during large or highly parallel workspace searches.
4. Slack: Best for Messages and Decision Context
Slack gives Claude access to conversations and decisions that may never have been documented elsewhere. Claude can search and retrieve messages, channels, threads, files and users, which makes it useful for finding old discussions or understanding how a decision was reached.
It also goes beyond search. Claude can draft and refine Slack messages, create and share canvases, and schedule messages to be posted later. Slack is currently listed as an Interactive connector.
Worth knowing: Claude must be installed in the Slack workspace before the Slack connector can be enabled. If your organization restricts Slack apps, an administrator may need to approve the installation first.
5. Linear: Best for Product and Engineering Workflows
Linear is useful when the work you discuss in Claude needs to end up in your actual backlog. Claude can search existing issues and projects, create new issues, update them, add comments and work with cycles. So you could give Claude a bug report, ask it to check Linear for related issues, then create or update the relevant issue without switching tools.
Worth knowing: Linear also offers a read-only MCP option. If you only want Claude to search and analyse your Linear workspace, you can restrict it from creating or changing anything.
6. GitHub: Best for Software Development Workflows
GitHub MCP is GitHub’s official MCP server and covers much more than repository search. Claude can read repositories and code, analyse commits, manage issues and pull requests, review code changes and search for specific code across a repository. The connector is currently listed as read and write.
That is why I wouldn’t describe GitHub as simply giving Claude “repository context.” It can participate in actual development workflows, from investigating code to reviewing a pull request or triaging issues.
Worth knowing: GitHub MCP is separate from Claude’s older Add from GitHub integration. That integration is designed to add selected repository files and folders to Claude as context rather than provide the broader MCP workflow capabilities described here.
7. HubSpot: Best for CRM and Revenue Workflows
HubSpot lets Claude query CRM information directly, including contacts, deals, companies and tickets, based on the connected user’s existing permissions. You can use it to analyse lifecycle progression, review win/loss patterns, inspect open tickets or answer questions using current customer and pipeline data.
The main value here is grounding Claude in real CRM information instead of relying on whatever deal or customer context you remember to include in the prompt.
If you want to take that a step further, check out our guide to CRM meeting integration with HubSpot for a closer look at how meeting data can feed into your CRM workflows.
Worth knowing: Claude’s current HubSpot connector is listed as Read, not read and write. I would therefore position it around CRM analysis and insights rather than claiming Claude can create or modify HubSpot records through this specific connector.
8. Asana: Best for Project and Task Management
Asana lets Claude turn planning into actual project work. It can search, create, update and track tasks, projects and goals through more than 30 tools connected to Asana’s Work Graph. You can ask Claude to find work due today, check a project’s status, create campaign tasks or turn a discussion into a project with tasks, owners and a timeline.
Asana is also an Interactive connector, so supported task and project experiences can appear directly inside Claude rather than being returned only as text.
Worth knowing: Claude can create and update real Asana work, so the permissions granted to the connected account matter.
9. Figma: Best for Design and Development Workflows
Figma works in both directions. Claude can use Figma designs, components, variables and design-system context to generate code, while supported workflows can also turn code into editable Figma designs. It can help maintain design systems, create FigJam diagrams and build content in Figma Slides.
This is another connector I wouldn’t reduce to a single use case. Its value sits across design handoff, implementation, design systems and the canvas itself.
Drawback: Usage depends heavily on your Figma seat and plan. View and Collab seats are limited to up to six read-tool calls per month, while Dev and Full seats receive higher limits. Write-to-canvas capabilities also have seat and permission requirements.
10. Canva: Best for Content and Design Creation
Canva lets Claude do considerably more than search for templates. Claude can browse and summarize Canva content, create new designs, edit designs, import files, export finished assets, resize content and work with supported Canva design tools directly from a conversation.
For a content team, this could be of real importance. Canva can become part of the production workflow rather than simply somewhere Claude sends you to find a design.
Drawback: Core features such as design generation, editing, search, exports and comments are available across Canva plans, but some tools are gated. Resizing requires Canva Pro or above, while template Autofill is currently Enterprise-only.
11. Stripe: Best for Payments and Revenue Analysis
Stripe is not just useful for analysing revenue data. Its Claude connector is currently read and write and gives Claude tools for interacting with the Stripe API as well as searching Stripe’s documentation and support knowledge. Claude can create and manage customers and create payment intents directly from a conversation.
That makes it useful for both finding information and completing supported Stripe operations without moving between Claude, documentation and the Stripe dashboard.
Worth knowing: Stripe’s MCP server is currently in public preview. It uses OAuth for granular, user-based authorization, and administrators can manage MCP access separately for sandbox and live environments.
12. Zapier: Best for Cross-App Automation
Zapier is the catch-all when Claude needs to work with an app that does not have its own connector, or when a workflow needs to cross several different tools. The connector is read and write and currently connects Claude with more than 8,000 apps through Zapier.
Claude can search connected apps and trigger actions such as sending emails, updating CRMs, scheduling meetings or turning information from a conversation into tasks elsewhere. tl;dv’s own integrations run through the same principle across 5,000-plus tools.
I’d use Zapier when the workflow genuinely spans multiple tools. If one direct connector already does everything you need, adding Zapier in the middle usually gives you an extra layer you do not need.
Drawback: Zapier MCP usage counts toward your Zapier task allowance, and that allowance is shared with your Zaps, AI steps, code and SDK usage. The free plan includes 100 tasks per month, so frequent Claude actions can burn through it quickly. Paid plans start at $19.99/month, with additional usage billed per task once you exceed your allowance.
Other Top Claude Connectors Worth Checking Out
These did not make the top twelve, but each is worth adding if it matches your stack.
| Connector | Категория | Add it when |
|---|---|---|
| Gmail | Общение | You want Claude drafting replies against real threads |
| Календарь Google | Производительность | Scheduling and prep across a busy week |
| Microsoft 365 | Производительность | Your work lives across SharePoint, OneDrive, Outlook and Teams |
| Airtable | Данные | Structured operational data lives outside your CRM |
| Ahrefs | Sales and marketing | SEO and AI search analytics in the chat |
| Salesforce | Sales and marketing | Enterprise CRM instead of HubSpot |
| Sentry | Код | Error triage alongside the repo |
| Amplitude | Данные | Product behaviour questions without a dashboard detour |
| Apollo.io | Sales and marketing | Prospecting and outbound research |
| Adobe Experience Manager | Производительность | Enterprise content operations |
Two of them carry a limit. Gmail drafts replies but cannot send them, and Microsoft 365 is read-only by default until an admin enables write tools, with Teams staying read-only either way.
How the Claude Connectors Directory Works
The directory lets you filter connectors by Works with, Use case, and Capabilities. The Capabilities filter is particularly useful because you can narrow the list to Read, Read & write, or Interactive connectors depending on what you want Claude to do. Connectors are also organized by use cases such as productivity, communication, data, design, code, and sales and marketing.
The Interactive label is worth knowing about. These connectors can render interfaces directly inside a Claude conversation instead of only returning text. That might be a task board, analytics dashboard, or design interface you can interact with without leaving the chat. Asana and Figma are two current examples.
The number of available connectors changes quickly. Anthropic reported more than 200 in April 2026 and more than 300 by June, so I wouldn’t rely on a fixed total in an article like this. Check the Claude Connectors Directory for the current list.
Claude Connector Limits and Plan Requirements
A few Claude connector limits are easy to miss:
- Free accounts get one custom connector – directory connectors are available across Claude plans, but Free users are limited to one custom connector added through a remote MCP URL.
- Ten connectors is where things can get crowded – Claude now uses Auto tool access by default and decides which connectors to load. Anthropic still recommends considering On demand when you have 10+ active connectors.
- Team and Enterprise plans gate connectors behind an Owner – an Owner or Primary Owner enables the connector, then members authenticate with the connected service individually. Owners can also restrict individual tools to Always allow, Needs approval, or Blocked. Enterprise-managed authentication is available in beta as an alternative for some organizations.
- Connectors only work in private projects – if a project is shared, connectors are not available there. Chats containing synced connector content also cannot be shared, and it is covered in more depth in Claude Projects.
Custom connectors using remote MCP connect to the server from Anthropic’s cloud rather than from your device, even when you’re using Claude Desktop or Claude Cowork. If the server sits behind a firewall or on a private network, you’ll need to allowlist Anthropic’s IP ranges before Claude can reach it.
Claude Connector Security: What to Check First
Before connecting anything, check what access the connector is actually requesting. The OAuth consent screen shows the permissions being requested from the connected service, while Claude’s own connector settings can separately control which tools are allowed to run.
Five checks you can do:
- Review the permissions during authentication – the connector page tells you what the integration is designed to do, but the authentication flow shows the access it is requesting from your account. For connected services, Claude can only access data your account already has permission to access.
- Use read-only access where it is available – if you only need Claude to search or analyse information, there is little reason to give it permission to create, edit or delete data. On Team and Enterprise plans, Owners can also block individual write/delete tools while keeping read access available.
- Be careful with “Allow always” – when Claude asks you to approve a connector tool, Anthropic recommends using this option only for servers and tools you trust to run without supervision. This matters even more for tools that can create, modify or delete data.
- Check how the connector provider handles your data – connected services process data on their own infrastructure and under their own terms and privacy policies. Their infrastructure may also be in a different region from Claude, and Claude’s inference-location settings do not control where third-party services process data.
- Disconnect connectors you no longer use – Anthropic specifically recommends removing services you no longer need. You can manage connected services and their permissions from Customize > Connectors.
None of this means connectors are inherently unsafe. The important thing is knowing what the connector can access, what actions it can take, and which company is processing the data before you grant it access.
If you are evaluating this for a team, tl;dv’s security commitment page is the kind of document worth asking every vendor for.
Which Claude Connectors to Add First
You don’t need to add every Claude connector at once, and I’d argue you shouldn’t. Every time I’ve set this up, adding one at a time and living with it for a week taught me more about what I actually needed than any upfront plan did. Start with one connector tied to work you do every week, then add the next only when you hit a real gap.
For most people, the first Claude connector to add is Google Drive, since it works on every plan and gives Claude the documents it’s usually missing. If your week runs on meetings the way mine does, I’d start there instead: connecting a meeting notetaker like tl;dv changed my recurring work rather than just a single afternoon, which is the case how to connect meeting notes to Claude MCP lays out in full, and it holds for any notetaker, not only ours.
From there, I add a connector only when a specific need shows up, never in advance. I reach for Slack when I keep needing the reasoning behind a decision, HubSpot when follow-ups need real deal data instead of invented context, and Linear when sprint work starts piling up. The habit that’s saved me the most hassle is stopping once Claude has enough context for the work in front of me, rather than collecting connectors I won’t touch. For a wider look at the meeting tools in that category, the best AI meeting assistants roundup covers where each one lands.
FAQs About Claude Connectors
Are Claude connectors free?
Directory connectors are available on every Claude plan, including free. Custom connectors are also available on free, but capped at one. The cost usually sits on the other side of the connection, since the connected service may need a paid tier of its own.
What is the difference between a Claude connector and an MCP server?
Every connector is an MCP server. Connector is the word for the packaged, hosted version you add in a click. MCP server is the broader term that also covers custom and locally run setups.
Can I use Claude connectors on mobile?
Yes. Web connectors work on Claude for iOS and Android, and connections you make on web or desktop appear on mobile next time you log in. Installing connectors on mobile is still in beta.
Do connectors work with Claude Code?
Yes. Connectors work across Claude, Claude Desktop, Claude Code, and the API. In Claude Code, local servers like tl;dv’s self-hosted option are registered from the command line with claude mcp add.
Which connector should a sales team add first?
The CRM, then the meeting connector, in that order. The CRM stops Claude inventing deal context, and the meeting connector supplies what was said on the call, which is the pairing behind most useful call follow-up. The AI notetaker for sales breakdown covers that workflow properly.
What happens when I disconnect a connector?
Claude loses access immediately and the tools disappear from your conversations. Nothing is deleted in the source service itself. Reconnecting means going through the consent screen again, which is also the fix when a connector’s permissions have changed and Claude stops seeing data it used to reach.
Is my data used to train Claude if I connect a service?
Connected services process data under their own terms, so the answer depends on the service. On tl;dv’s side, the privacy policy states that Customer Content, including recordings, transcripts, and notes, is not used to train, fine-tune, or improve foundation models, large language models, or other generative AI models, for tl;dv or any third party.
