14 Best Claude Connectors That Earn a Spot in Your Daily Workflow (2026)

TL;DR: Best Claude connectors

• Firecrawl: Live web search, scraping, crawling, monitoring, and structured provider data
• Asana: Turn conversations into projects, tasks, and status checks
• Lovable: Build, iterate on, and deploy full apps from a Claude chat
• Granola: Ask questions across your meeting notes and transcripts
• tldraw: A live, editable canvas Claude draws diagrams on
• Notion: Search and write pages and databases
• Slack: Search threads, summarize channels, send messages
• Google Drive: Find and read Docs, Sheets, Slides, and PDFs
• Linear: File and update issues, projects, and cycles
• HubSpot: CRM records, pipeline, and campaign analytics
• Hex: Warehouse questions answered by the Hex agent
• Figma: Design context for code, tokens, and FigJam diagrams
• Canva: Generate and export on-brand designs
• Stripe: Billing records, revenue analytics, and Stripe docs

Without connectors, Claude works from memory. With connectors, it can check the live web, file the task, read last Tuesday's call, and ship the prototype in the same conversation, and that's where most of my time savings in 2026 have come from.

The connector directory has grown fast since Anthropic opened it up, and most entries are thin wrappers that you install once and forget. Anthropic also keeps building on top of connectors. On October 8 it launched Claude Dashboards in beta, which builds live dashboards from a data platform like BigQuery or Snowflake and works with your other connectors too. The same release moved Docs, Slides, and Design onto every plan, including Free, so whatever a connector pulls in can end up as a finished doc or deck.

Like a lot of people, I spend about 90% of my workday in Claude, so I've tested a lot of connectors and kept the fourteen that save me the most time. These are the best Claude connectors I'd hand someone starting today.

A Claude connector is an integration that lets Claude call tools on an outside service while you chat. Under the hood every connector is an MCP server. The service publishes a list of tools (search, create task, read transcript, draw shape), and Claude decides when to call them based on your prompt.

There are three ways to get one into your setup:

• Directory connectors. Listed in Claude's connector browser at claude.ai/directory. Click Connect and sign in with OAuth. Every pick on this list lives here.
• Custom connectors. Any service with a remote MCP URL. Go to Customize, Connectors, Add custom connector, paste the URL, and sign in.
• Claude Code. claude mcp add --transport http <name> <url> adds a remote MCP server from the terminal, and /mcp finishes the sign-in. Some vendors also ship plugins that bundle the connector with slash commands and skills; see the best Claude Code plugins roundup for those.

Some connectors are MCP Apps, which return an interactive UI instead of plain text. tldraw is one, and it renders a full drawing canvas inside the chat.

For any directory connector:

1. Open the connector picker. In a Claude chat, click the + icon, then Connectors, then Browse Connectors.

2. Find and connect. Search for the service (Firecrawl, Asana, Notion, Slack, and so on), click Connect, and complete the sign-in in the browser window that opens.

3. Switch it on. Back in the chat, open +, go to Connectors, and toggle the connector on. You can mention it by name in a prompt ("Use Granola to...") or let Claude pick it when the prompt matches.

On Claude Team and Enterprise plans, an owner may need to approve a connector before members can add it. Early-stage companies may get Team free for a year. The expanded Claude for Startups program now offers qualifying startups a free year of Claude Team with up to five premium seats plus $1,000 in API credits, per TechCrunch. Companies founded in the last five years or funded in the last two qualify.

The Firecrawl connector gives Claude live, clean web data: search with full page content, scraping, crawling, browser interactions, monitoring, and structured data from official providers.

Claude's built-in web tools return snippets and short fetches. That's fine for a quick fact, but it breaks down the moment you need the full page, a JavaScript-rendered pricing table, a 40-page PDF, or the same data from 200 URLs. Firecrawl handles those cases, and the other connectors on this list get more useful once Claude can read full pages.

It also gives Claude data that its built-in tools can't reach at all. Through Firecrawl Alexandria, Claude pulls structured records directly from data providers, such as funding rounds, stock news, or verified work emails, instead of hoping a web page quoted the right number. The Claude web fetch vs Firecrawl comparison covers where web fetch stops and which providers Claude can reach through Alexandria.

Firecrawl is one of the top 50 most-starred repos on GitHub and powers web data for 150,000+ companies. Output comes back as LLM-ready markdown or JSON, which keeps Claude's context window clean even on large jobs.

What Claude gets access to:

• search: Web search that returns the full content of matching pages alongside the links
• scrape: Clean markdown or structured JSON from any URL, including JavaScript-heavy sites, PDFs, and DOCX files
• crawl: Follow links across an entire site to pull a docs section or build a dataset
• interact: Act on a page after scraping it: click, fill forms, paginate, and pull dynamic content
• monitor: Watch pages for changes (pricing pages, changelogs, job boards) and get alerted
• parse: Turn PDFs, DOCX, XLSX, and PPTX into clean markdown
• agent: Describe the data you need and let Firecrawl find and extract it across many sites
• Firecrawl Alexandria: Firecrawl's catalog of data providers plus its Research and Developer indexes. Claude finds a provider tool, checks its inputs and price, and runs it for structured records without scraping. Finding tools is free; each run costs the tool's listed price

Install:

In Claude, click +, then Connectors, then Browse Connectors, search for Firecrawl, and click Connect. Sign in, pick the team to bill, and approve. New users can create a free account at that step. The Firecrawl in Claude page walks through it.

For Claude Code, install the official plugin:

# Official plugin (adds /firecrawl:* slash commands)
claude plugin install firecrawl@claude-plugins-official
 
# Or add the remote MCP server directly
claude mcp add firecrawl --transport http https://mcp.firecrawl.dev/YOUR_API_KEY/v2/mcp

Example prompts:

"Search the web for this week's AI infrastructure launches and give me a briefing with sources"
 
"Scrape the pricing pages of our three closest competitors and put the tiers in a table"
 
"Crawl docs.stripe.com/webhooks and summarize how to verify event signatures"
 
"Monitor https://openai.com/news and tell me when a new post goes up"
 
"Parse this 60-page annual report and pull the revenue by segment into JSON"
 
"Find the latest analyst upgrades on Nvidia from an official provider"

Honest take: This is the first connector I install on any Claude account. The others on this list each handle one job, while Firecrawl fixes Claude's limited view of the live web, which affects most of them. With it on, Claude reads the current pricing page or docs instead of guessing.

I use Alexandria more than I expected, mostly for Firecrawl's research paper and developer docs indexes alongside the data providers.

Cons: Heavy crawls and Alexandria provider runs use credits, so large jobs need a paid plan. Some Alexandria providers require an admin to accept the provider's data terms once in the Firecrawl dashboard before Claude can run them.

Connector page: claude.ai/directory/connectors/firecrawl. Docs at docs.firecrawl.dev.

The Asana connector brings Asana's Work Graph into Claude so you can search, create, update, and track tasks, projects, portfolios, and goals from a conversation.

Planning with Claude used to end with someone copying the outcome into a project tracker, and the Asana connector removes that step. Claude can turn a brief into a project with sections and tasks, assign work, and answer "what's at risk this week?" across your portfolios.

Asana lists 30+ tools behind the connector, including task and project creation, portfolio lookups, status overviews, and workspace search.

What you get:

• create_project_preview / create_task_preview: Draft projects and tasks from the conversation, with a preview before anything is created
• update_task: Reassign, reschedule, or complete existing tasks
• get_status_overview: Status across projects and portfolios in one call
• get_portfolios / get_items_for_portfolio: Portfolio-level reporting from chat
• search_objects: Find tasks, projects, and people across the workspace

Install:

In Claude, click +, choose Connectors, search for Asana, click Connect, sign in, and grant permissions. Then toggle Asana on from the Connectors menu.

For Claude Code or other MCP clients, Asana's V2 server lives at:

claude mcp add --transport http asana https://mcp.asana.com/v2/mcp

Example:

"Turn this launch brief into an Asana project with milestones for design, engineering, and marketing"
 
"What campaigns in the Q4 Marketing portfolio are at risk of missing their dates?"
 
"Assign every copywriting task in the Onboarding Revamp project to the Content team and set them due next Friday"

Honest take: I keep this on for the write actions, because turning a planning chat into tasks was always the slow part. The preview-then-create flow is a sensible guardrail too, since you see the task breakdown before it lands in a shared project.

Claude tends to over-generate tasks on a first pass, so I usually trim a few before confirming.

Cons: On a Claude Team account, Asana requires organization-wide app approval, which needs App Management; that isn't included in Asana's Starter plan, so Starter teams can get stuck. Some users have reported tools briefly disappearing from the connector after updates, which a reconnect usually fixes.

Connector page: claude.com/connectors/asana. Reference: Asana app in Claude.

The Lovable connector lets Claude create, iterate on, inspect, and deploy full Lovable apps without leaving the chat.

Lovable turns a prompt into a working web app. With its MCP server, you stay in Claude and Claude drives Lovable for you. It creates the project, sends follow-up build messages, reads the code, checks the diff, and publishes when you're happy.

That makes Claude a good front end for app building. You can research a market with Firecrawl, write the spec with Claude, and hand it to Lovable in the same conversation. If you want to go deeper on pairing the two, the Firecrawl and Lovable tutorial shows how to wire live web data into a Lovable app.

What you get:

• create_project: Start a new app from a build prompt, optionally from a template or design system
• send_message: Keep iterating ("add a dark mode toggle", "make the pricing page mobile-friendly")
• get_diff / list_files / read_file: Inspect exactly what Lovable changed
• deploy_project: Publish to lovable.app and get the live URL back
• enable_database / query_database: Turn on Lovable Cloud and run SQL against it
• import-claude-design-from-url: Claude-only tool that imports a Claude design into a new Lovable project

Install:

In Claude, click +, then Connectors, then Browse Connectors, search for Lovable, click Connect, and log in to Lovable.

# Claude Code: add the MCP server
claude mcp add --transport http lovable "https://mcp.lovable.dev"
 
# Or install the plugin, which adds slash commands for common tasks
/plugin install lovable@claude-plugins-official

Example:

"Create a Lovable project called Feedback Hub with a form for collecting user feedback and an admin view"
 
"Build three versions of a landing page for this product, each with a different layout, deploy all three, and give me the URLs"
 
"Show me what changed in the last three edits to my project"

Honest take: Building three landing page variants and getting three live URLs back from one prompt is the moment this connector clicked for me. I also use it to review work, since asking Claude to read the diff and explain what Lovable changed is faster than clicking through the editor. It works best when you already know roughly what you want, because vague prompts burn credits on rebuilds.

Cons: The connector has the same access you do across your whole Lovable account, and query_database runs with full read, write, and schema permissions. create_project and send_message use standard Lovable credits. On Free and Pro plans, anything you deploy is public to anyone with the link. On Enterprise workspaces, an admin has to enable third-party MCP clients first.

Connector page: claude.com/connectors/lovable. Docs: docs.lovable.dev/integrations/lovable-mcp-server. Repo: github.com/lovablelabs/mcp.

The Granola connector gives Claude your meeting history: search notes, pull action items, and quote transcripts in any conversation.

Granola is an AI notepad that captures meetings without a bot joining the call. Its first-party MCP connector means Claude can answer "what did we agree with Acme last week?" without me digging through notes or pasting transcripts.

It also combines well with the rest of the list. In one prompt, Claude can pull the action items from a call, research the companies mentioned with Firecrawl, and file the follow-ups in Asana.

What you get:

• query_granola_meetings: Ask natural-language questions across your meeting history
• list_meetings: Find meetings by title, date, or attendee
• get_meetings: Read full notes, including your private notes and the AI summary
• get_meeting_transcript: Pull the raw transcript for exact quotes (paid plans)
• list_meeting_folders: Browse meeting folders (paid plans)

Install:

In Claude, go to Settings, then Connectors, search for Granola, click Connect, and authenticate. Make sure it's toggled on in the chat.

# Claude Code
claude mcp add granola --transport http https://mcp.granola.ai/mcp
# then run /mcp, select granola, and choose Authenticate

Example:

"Using my Granola notes from the Acme call on Tuesday, draft a follow-up email with the open questions"
 
"What are the remaining action items for each person on the pricing project?"
 
"Give me exact quotes from this week's customer calls where someone described onboarding as confusing"

Honest take: This is the connector I'd recommend to anyone who spends half their week in meetings. Meeting notes are the context Claude is most often missing, and Granola fills that gap without any manual export. The transcript tool is the best part, because Claude can quote what someone said instead of paraphrasing a summary.

Cons: The free Basic plan only covers the last 30 days of notes, and transcripts and folder browsing are paid-plan tools. MCP only sees your active Granola workspace, so if you have personal and work workspaces you have to switch in the Granola app first. Rate limits sit around 100 requests per minute.

Connector page: claude.com/connectors/granola. Docs: docs.granola.ai/help-center/sharing/integrations/mcp.

The tldraw connector puts a live, editable tldraw canvas inside your Claude chat so Claude can sketch flowcharts, architecture diagrams, wireframes, and mind maps with you.

Some ideas are hard to explain in text: a system architecture, a user flow, a layout tweak. tldraw's connector is an MCP App, which means it returns an interactive canvas rather than a description of one. Claude draws on it, you edit it, and Claude sees your changes and keeps building.

tldraw itself is an open infinite-canvas SDK with 50.8K GitHub stars, and the connector is built on that same SDK.

What you get:

• create_shapes: Draw boxes, arrows, text, and freeform shapes
• update_shapes: Move, restyle, or relabel what's already on the canvas
• delete_shapes: Clean up as the diagram evolves
• save_checkpoint / read_checkpoint: Save the canvas state and pick it back up later

Install:

In Claude (web or desktop), go to Settings, Connectors, Browse Connectors, search for tldraw, and click Connect. Open a chat and ask it to draw something.

# Claude Code or other MCP clients
claude mcp add tldraw --transport http https://tldraw-mcp-app.tldraw.workers.dev/mcp

Example:

"Draw a diagram of our signup flow: landing page, OAuth, onboarding survey, and first API call"
 
"Map out the architecture for a RAG pipeline that uses Firecrawl for ingestion"
 
"Wireframe a pricing page with three tiers and a comparison table underneath"

Honest take: Two-way editing is why I use this over asking Claude for a Mermaid diagram. I can drag a box to where it belongs, add an arrow Claude missed, and ask it to revise the plan based on the updated drawing. For explaining a system to a teammate, it's become my default.

Cons: Claude's layouts are functional but not pretty, so diagrams headed for docs or a deck still need a visual pass. It's most useful in clients that render MCP Apps (Claude web, Claude Desktop, Cursor). For pixel-level UI work you'll still want Figma, and if you want the idea to move, Anthropic's new Claude Motion (beta, Team and Enterprise) turns a report or chart into an animated explainer you can export as an MP4.

Connector page: claude.com/connectors/tldraw. Reference: tldraw.dev/blog/text-is-not-enough. Repo: github.com/tldraw/tldraw.

The Notion connector lets Claude search, read, create, and update pages and databases in your Notion workspace.

If your team's wiki, specs, and meeting docs live in Notion, this is the connector that gives Claude the background it usually lacks. Claude can find the right page without you pasting a link, and it can write results back as a new page or a database entry instead of leaving them in the chat.

What you get:

• search / fetch: Find and read pages and databases across the workspace
• create-pages / update-page: Write new docs or edit existing ones
• create-database / update-database: Build and change structured trackers
• create-comment / get-comments: Leave and read comments on pages

Install:

In Claude, click +, then Connectors, then Browse Connectors, search for Notion, click Connect, pick the workspace, and approve access.

# Claude Code: add Notion's remote MCP server
claude mcp add --transport http notion https://mcp.notion.com/mcp

Notion also publishes a Claude Code plugin that adds the server plus skills and slash commands.

Example:

"Find our Q4 launch spec in Notion and list the open questions nobody has answered"
 
"Create a Notion page summarizing this research, with a table of the five competitors and their pricing"
 
"Add a row to the Content Calendar database for each post idea in this chat"

Honest take: Notion is where I send Claude's output when it needs to outlive the conversation. Pairing it with Firecrawl works well, since Claude researches a topic on the live web and files a clean summary page in the right folder.

Cons: Claude only sees pages your Notion account can access, so shared-but-unjoined teamspaces won't show up. Large workspaces return a lot of loosely related search hits, so naming the page or database in your prompt gets better results.

Connector page: claude.com/connectors/notion. Docs: developers.notion.com/guides/mcp.

The Slack connector lets Claude search messages, read channels and threads, send messages, and create canvases in your workspace.

Decisions get made in Slack threads and then forgotten. With the connector on, Claude can find where something was decided, summarize a long thread, and draft or send the follow-up in the same conversation.

What you get:

• slack_search_public / slack_search_public_and_private: Search messages across channels
• slack_read_channel / slack_read_thread: Pull full channel history or a single thread
• slack_send_message: Post a message or reply
• slack_create_canvas / slack_update_canvas: Write longer summaries as Slack canvases
• slack_search_users / slack_read_user_profile: Look up who's who

Install:

In Claude, click +, then Connectors, then Browse Connectors, search for Slack, click Connect, select the workspace, and approve the requested permissions.

# Claude Code: Slack's official plugin configures the MCP server and OAuth
claude plugin install slack

Example:

"Summarize everything in #product-launches from the last week and list what still needs a decision"
 
"Find the thread where we agreed on the new pricing tiers and quote the final numbers"
 
"Turn this research into a canvas in #growth and post a two-line summary linking to it"

Honest take: Thread summaries alone justify this one after a few days away. Being able to send the message from Claude saves the copy-paste step, though I still read anything addressed to a person before it goes out.

Cons: A workspace admin has to approve the Slack MCP integration before anyone can connect, which can take a while at larger companies. Private-channel search depends on the scopes you grant.

Connector page: claude.com/connectors/slack. Docs: docs.slack.dev/ai/slack-mcp-server/connect-to-claude.

The Google Drive connector lets Claude search your Drive, read Docs, Sheets, Slides, and PDFs, and create or upload files.

Most of my working documents live in Drive, and this connector means I stop downloading a file only to upload it to Claude again. Claude can find a document when I only remember roughly what it was about, read the current version, and save its output back to Drive.

Anthropic also announced separate Google Docs, Sheets, and Slides connectors in beta on October 6, along with an add-on that brings Claude into those files directly. Drive is still the one that searches across everything.

What you get:

• search_files / list_recent_files: Find documents by content or recency
• read_file_content: Read Docs, Sheets, Slides, and PDFs
• create_file / update_file / copy_file: Write new files or edit existing ones
• share_file / get_file_permissions: Check and change who can see a file

Install:

In Claude, click +, then Connectors, then Browse Connectors, search for Google Drive, click Connect, and sign in with your Google account.

Example:

"Find the pricing analysis spreadsheet from last month and tell me which plan had the highest churn"
 
"Read the three customer interview docs in my Research folder and pull out the common complaints"
 
"Save this competitor summary as a new Google Doc in my Drive"

Honest take: It's the least exciting connector on this list and one of the ones I use most. Combined with Firecrawl, Claude can compare what's on a competitor's live site with what our internal doc says about them.

Cons: Search quality depends on how well your files are named and organized. Sharing and trashing files are real write actions, so check what Claude proposes before approving them.

Connector page: claude.com/connectors/google-drive.

The Linear connector lets Claude read, create, and update issues, projects, cycles, and comments in Linear.

Planning chats often end with a list of bugs or follow-ups. With Linear connected, Claude files them as real issues with the right team, labels, and status, and it can pull in what's already in the backlog before you plan.

What you get:

• list_issues / get_issue / list_my_issues: See what's open and what's assigned to you
• create_issue / update_issue: File new issues or change status, assignee, and labels
• list_projects / create_project / update_project: Manage projects
• list_cycles: Check what's in the current cycle
• create_comment / list_comments: Discuss issues without opening Linear

Install:

In Claude, click +, then Connectors, then Browse Connectors, search for Linear, click Connect, and complete OAuth.

# Claude Code
claude mcp add --transport http linear-server https://mcp.linear.app/mcp
# then run /mcp and complete OAuth

Example:

"File a Linear issue for each bug in this Slack thread, assign them to the Platform team, and label them P1"
 
"What's left in the current cycle for the Growth team, and what's at risk?"
 
"Summarize every issue I closed this week for my standup notes"

Honest take: For engineering teams this replaces Asana's role on the list. The issue-filing flow is fast, and Claude is good at writing a proper reproduction section when you give it the context.

Cons: Claude sometimes picks the wrong team or label on a first pass in larger workspaces, so name them in the prompt. Like any write-capable connector, it can create a lot of issues quickly if your prompt is loose.

Connector page: claude.com/connectors/linear.

The HubSpot connector brings CRM context into Claude: search and update contacts, companies, deals, and tickets, and analyze campaign performance.

Sales and marketing work in Claude goes better when Claude knows who the customer is. The HubSpot connector gives it the deal history, pipeline stage, and campaign data so call prep, account updates, and pipeline reviews are grounded in the actual record.

HubSpot lists 36 tools behind the connector, covering CRM objects, marketing emails, landing pages, blog posts, and campaign attribution.

What you get:

• get_crm_objects / manage_crm_objects: Read and update contacts, companies, deals, and tickets
• query_crm_data: Ask questions across your CRM in plain language
• get_campaign_attribution_reports: See which campaigns drove pipeline
• get_marketing_email_analytics / get_content_analytics_report: Email, page, and blog performance
• manage_landing_page / manage_blog_post: Draft and edit marketing content

Install:

In Claude, click +, then Connectors, then Browse Connectors, search for HubSpot, click Connect, and sign in to your HubSpot portal. Grant only the scopes you need.

Example:

"Prep me for my call with Acme: deal stage, last three touchpoints, and open tickets"
 
"Which campaigns this quarter influenced the most closed-won revenue?"
 
"Move every deal that's had no activity in 30 days to the Stalled stage and list them"

Honest take: Pair this with Firecrawl and Claude can enrich an account with what's on the company's site today, then update the CRM record in the same pass. That's the workflow I'd set up first.

Cons: The connector needs a paid Claude plan (Pro, Max, Team, or Enterprise). For team-wide rollouts, a HubSpot super admin has to install and configure it first, and existing users must reconnect when HubSpot adds new scopes.

Connector page: claude.com/connectors/hubspot. Reference: HubSpot connector for Claude.

The Hex connector lets Claude search your Hex projects and ask data questions through the Hex agent, which queries your warehouse and returns charts and analysis.

When a question needs real numbers from the warehouse, Claude hands it to the Hex agent, which writes the SQL, runs it, and returns the analysis. It also fits Anthropic's new Claude Dashboards, which can send a dashboard straight to Hex when a question needs deeper analysis.

What you get:

• search_projects: Find existing Hex projects by name or content
• create_thread: Start a Hex agent thread with a data question
• get_thread / continue_thread: Read results and ask follow-ups

Install:

In Claude, click +, then Connectors, then Browse Connectors, search for Hex, click Connect, and authorize your Hex workspace.

Example:

"Ask Hex how weekly signups this month compare with last month, split by acquisition channel"
 
"Find the Hex project on churn analysis and summarize its latest findings"
 
"What share of new accounts in September made a second API call within seven days?"

Honest take: Our team runs on Hex, so this one gets used a lot. Answers come back with the query and charts attached, which makes them easy to check before I repeat a number to anyone.

Cons: Hex's connector is available on Team and Enterprise plans, and users need the Explorer workspace role or higher. Complex agent threads can take several minutes to finish.

Connector page: claude.com/connectors/hex. Docs: learn.hex.tech/docs/api-integrations/mcp-server.

The Figma connector gives Claude design context from your Figma and FigJam files so it can generate code that matches the design, create diagrams, and pull design tokens.

If you build front ends with Claude, this closes the gap between "here's a screenshot" and "here's the actual frame." Claude reads the layout, variables, and component mappings straight from the file, which produces code much closer to the design than a description would.

What you get:

• get_design_context: Pull structured context for a frame or component to generate code from
• get_screenshot / get_metadata: See the design and its layer structure
• get_variable_defs: Read colors, spacing, and typography tokens
• get_code_connect_map: Map Figma components to components in your codebase
• generate_diagram / get_figjam: Create and read FigJam diagrams

Install:

In Claude, click +, then Connectors, then Browse Connectors, search for Figma, click Connect, and approve access.

# Claude Code: Figma's hosted MCP server
claude mcp add --transport http figma https://mcp.figma.com/mcp

Example:

"Build this Figma frame as a React component using our existing Button and Card components"
 
"List the color and spacing variables used in this file"
 
"Create a FigJam flowchart of our onboarding steps"

Honest take: For design-to-code it's the most useful connector I've tried, because Claude works from real tokens and structure. It overlaps a little with tldraw for diagrams, but Figma is where finished design work lives.

Cons: Rate limits depend heavily on your Figma seat. Starter plans and View or Collab seats get only 6 tool calls a month, so you realistically need a Dev or Full seat on a paid plan.

Connector page: claude.com/connectors/figma. Docs: Guide to the Figma MCP server.

The Canva connector lets Claude search, generate, autofill, and export Canva designs from the chat.

For marketers who don't work in Figma, Canva is the quickest way from a written brief to a finished social post or deck. Claude can generate a design, find an existing one to reuse, and export it without you opening Canva.

What you get:

• generate-design: Create a new design from a prompt
• search-designs / get-design-content: Find and read existing designs
• import-design-from-url: Bring a file into Canva from a link
• export-design: Export to PDF, PNG, and other formats
• add-comment-thread-to-design: Leave feedback on a design

Install:

In Claude, click +, then Connectors, then Browse Connectors, search for Canva, click Connect, and sign in to your Canva account.

Example:

"Design a LinkedIn carousel from this blog post using our brand colors"
 
"Find last month's webinar deck in Canva and export it as a PDF"
 
"Generate three Instagram post options announcing our new feature"

Honest take: It covers most everyday marketing graphics without leaving Claude. Now that Claude's own Slides and Design are on every plan, I use Canva mainly when the team's brand kit and templates already live there.

Cons: Output follows Canva's template conventions, so highly custom brand work still needs manual cleanup. Multi-page designs sometimes need fixes in the Canva editor.

Connector page: claude.com/connectors/canva. Reference: Connect Canva to an AI assistant.

The Stripe connector gives Claude access to your Stripe account and docs: look up customers, invoices, subscriptions, and balances, run analytics, and search Stripe's documentation.

When a support ticket or a revenue question involves billing, Claude can check the actual Stripe records instead of guessing. The same connector also searches Stripe's docs, which helps when you're building on Stripe.

What you get:

• stripe_api_search / stripe_api_read: Find and read customers, invoices, subscriptions, and charges
• stripe_api_write: Create or update Stripe objects
• get_balance_summary / stripe_analytics: Balances and revenue analytics
• search_stripe_documentation: Search Stripe's docs and support articles
• stripe_implementation_planner: Plan a Stripe integration

Install:

In Claude, click +, then Connectors, then Browse Connectors, search for Stripe, click Connect, and authorize with OAuth.

# Claude Code: Stripe's hosted MCP server
claude mcp add --transport http stripe https://mcp.stripe.com

Example:

"Why did this customer's last payment fail, and when is their next retry?"
 
"How much MRR did we add from annual plans in September?"
 
"Look up how to handle proration when a customer upgrades mid-cycle"

Honest take: For anyone running a SaaS product, this is the connector that turns billing questions into a single prompt. Docs search alone is handy for integration work.

Cons: Stripe's MCP server is in public preview. stripe_api_write can change real billing data, so I'd use test mode or a restricted key while you get comfortable with it.

Connector page: claude.com/connectors/stripe. Docs: docs.stripe.com/mcp.

Every connector you switch on adds tool definitions Claude has to read, so keep the always-on set small and toggle the rest per chat.

My everyday set is Firecrawl for anything that touches the live web, Granola for meeting context, and Asana to turn both into tracked work. Lovable and tldraw come on when I'm building or explaining something, and Hex and HubSpot when the question involves our own numbers.

These connectors also chain well. A typical run looks like "Pull the action items from today's Granola call, research the two competitors that came up with Firecrawl, sketch the proposed architecture in tldraw, and create Asana tasks for each next step." That used to be four tabs and an afternoon. With Docs and Slides now on every plan, the last step can be "turn this into a deck for Monday" without leaving the chat.

For discovery, claude.ai/directory is the official list of top Claude connectors, and any remote MCP server can go in as a custom connector. If you work in the terminal, the best MCP servers for Claude Code guide has claude mcp add commands for each server, and the best Claude Code skills roundup covers the skills that pair with them. If you also use ChatGPT, the best ChatGPT plugins list covers the same ground there, and scheduling tasks in Claude Desktop with Firecrawl shows how to put connectors on a timer.

If you install one connector from this list, make it Firecrawl, because every other connector works with whatever Claude can see.