ScreenshotNeo
Sign in Get a free key

BlogAI agents

Most Popular MCP Servers for Claude Code

A practical, documented shortlist of MCP servers for Claude Code, with setup commands, security guidance, and workflow examples.

By the ScreenshotNeo team1 October 20268 min read

Short answer: There is no published adoption ranking for Claude Code MCP servers. The most useful shortlist is the set of servers Anthropic documents as commonly used, organized by the work they enable: development, project management, data, payments, design, infrastructure and automation. Choose by task, permissions, transport and trust rather than an unverified popularity score.

Model Context Protocol (MCP) is an open standard that lets AI applications connect to external tools, data sources and workflows. In Claude Code, an MCP server can let the assistant inspect monitoring data, work on issues and pull requests, query a database, read design files or automate follow-up work without requiring you to paste all the context into chat. Anthropic’s Claude Code MCP reference lists examples it calls commonly used, but it does not publish server-by-server usage counts or a ranking method.

Popularity here means officially documented or commonly referenced examples, not measured market share. Repository stars, search results and anecdotal usage are poor substitutes for a dated, transparent methodology, so this guide does not claim a numbered top-ten ranking. Availability, pricing and authentication can change; confirm current details in each provider’s documentation before connecting an account.

Quick comparison by job

Job Documented MCP examples Typical Claude Code task
Development and testing Sentry, Socket, Hugging Face, Jam Investigate production errors, inspect dependency risk, query Hub or Gradio information, review debugging recordings
Project management and documentation Asana, Linear, Atlassian, Notion, ClickUp, Monday, Intercom, Fireflies Find issues, update tickets, read Confluence pages, summarize customer conversations or meetings
Databases and data Airtable, Daloopa, HubSpot and other database or CRM servers Retrieve records, inspect structured data and prepare follow-up work
Payments and commerce PayPal, Plaid, Square, Stripe Look up payment context, investigate transactions or build commerce workflows
Design and media Figma, Cloudinary, Canva Read design handoff data, find media assets and coordinate changes
Infrastructure and deployment Cloudflare, Netlify, Vercel Inspect projects, deployments, logs and documentation
Automation Workato, Zapier Trigger multi-step actions across connected systems

Development and testing

  • Sentry: useful for error monitoring and debugging production issues.
  • Socket: focused on dependency security analysis.
  • Hugging Face: provides Hub information and access to Gradio applications.
  • Jam: exposes debugging recordings and diagnostics.

These servers are most valuable when a coding task depends on evidence outside the repository. Limit access to the projects and environments the assistant actually needs.

Project management and documentation

  • Asana and Linear: issue and project work.
  • Atlassian: Jira and Confluence workflows.
  • Notion: pages and tasks.
  • ClickUp and Monday: task management.
  • Intercom and Fireflies: customer conversations and meeting transcripts.

A common pattern is: ask Claude Code to find the relevant issue, inspect the repository, propose a change, then write a concise status update back to the project system. Keep write operations behind explicit confirmation in your team’s process.

Databases, CRM and business data

Airtable, Daloopa, HubSpot and other database or CRM integrations can supply records that are otherwise difficult to copy into a coding session. Start with read-only credentials and narrow schemas or workspaces. Avoid granting broad production write access for a task that only requires a lookup.

Payments and commerce

PayPal, Plaid, Square and Stripe are documented examples for payment-related workflows. Treat payment data as sensitive: use a test account where possible, restrict scopes and avoid placing secrets or full customer records in prompts or logs.

Design and media

Figma, Cloudinary and Canva connect implementation work to design files and media assets. They are useful for checking measurements, locating the approved asset and confirming that an implementation matches the handoff.

Infrastructure and deployment

Cloudflare, Netlify and Vercel support infrastructure and deployment workflows. Anthropic describes Vercel’s official server as supporting documentation search, projects and deployments, and deployment-log analysis. Use a read-only role for diagnosis, then a separate, deliberately approved path for production changes.

Automation

Workato and Zapier can connect Claude Code to multi-step business automations. Because one tool call may trigger several downstream actions, document the workflow and require confirmation before any irreversible step.

How to add an MCP server to Claude Code

Anthropic recommends HTTP for remote servers and calls it the most widely supported transport for cloud services. Use local stdio when a process needs direct access to the machine. Remote WebSocket servers are appropriate for services that push events. SSE is deprecated; use HTTP when a provider supports it.

1. Add a remote HTTP server

claude mcp add --transport http notion https://mcp.notion.com/mcp

The service may open an authentication flow or require an access token. Follow the provider’s current instructions rather than guessing an endpoint or scope.

2. Add a local stdio server

claude mcp add my-local-server -- npx -y @example/mcp-server

Everything after -- is the local command Claude Code starts. Pin a known package version in team environments and review what the process can read from disk.

3. Choose a scope

Use a personal or local scope for experiments. Use project scope when a team should share the configuration. A checked-in .mcp.json makes the project setup reproducible, but review it like code because it defines external access.

4. Inspect and verify the connection

claude mcp list
claude mcp get notion

Run a harmless read operation first. Confirm the account, workspace, project and permissions before asking Claude Code to create, modify or delete anything.

Example project configuration

{
  "mcpServers": {
    "notion": {
      "type": "http",
      "url": "https://mcp.notion.com/mcp"
    },
    "local-tools": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "@example/mcp-server"]
    }
  }
}

The exact fields and authentication mechanism depend on the server. Treat this as the shape of a project configuration, not a universal manifest for every provider.

How to choose the right server

  1. Define the job. Decide whether you need issue management, observability, design handoff, data access or deployment.
  2. Map the data. List the workspaces, repositories, projects and records Claude Code must reach.
  3. Check authentication. Prefer OAuth or scoped tokens, and separate read and write credentials.
  4. Check transport. Prefer HTTP for hosted services; use stdio for local processes and WebSocket when event delivery is required.
  5. Decide who owns setup. Personal setup is quicker; project setup is repeatable for a team.
  6. Test the smallest useful operation. Read one issue, page or deployment before enabling mutations.

Security and trust checklist

Anthropic explicitly advises: “Verify you trust each server before connecting it.” Its catalog also warns that not every third-party server has been verified for correctness or security. A server that fetches external content can expose Claude Code to prompt injection.

  • Confirm who operates the server and read its current documentation.
  • Review source code or vendor security material when available.
  • Grant the smallest useful permission set and workspace.
  • Keep production credentials separate from development credentials.
  • Assume external content may contain instructions aimed at the model.
  • Require human review before destructive actions, deployments or bulk updates.
  • Rotate tokens when a project member, machine or integration changes.

Performance, reliability and cost considerations

  • Latency: remote calls add network and authentication time. Ask for focused fields instead of entire workspaces.
  • Reliability: a failed server can block a workflow. Keep a manual fallback and capture request IDs or provider error details.
  • Context size: large issue lists, transcripts and database results consume model context. Filter by project, date, status or identifier.
  • Rate limits: respect the provider’s limits and batch only where its API supports safe batching.
  • Cost: Claude Code usage and the connected service may each have separate pricing. Check both providers’ current plans; this guide does not infer a combined price.
  • Availability: an official server can still be unavailable or require a changed authentication flow. Keep configuration and credentials documented.

Troubleshooting common errors

Symptom Likely cause Fix
Server does not appear in Claude Code Wrong scope, malformed command or configuration not reloaded Run claude mcp list, inspect the project file and restart Claude Code after changes.
401 or 403 response Expired token, wrong account or insufficient scope Re-authenticate, verify the workspace and request the minimum required permission.
Connection timeout Unavailable endpoint, proxy or network policy Check the provider status and endpoint, then test from the same machine and network.
stdio process exits immediately Missing runtime, package, environment variable or incompatible version Run the command manually, pin the package version and provide required environment variables.
Claude receives irrelevant results Query is too broad or the server exposes multiple workspaces Specify project, repository, date range and record type in the request.
Unexpected tool action Overly broad permissions or ambiguous instructions Reduce scopes, switch to read-only credentials and require confirmation for writes.
Prompt-injection warning Fetched content contains instructions directed at the model Treat the content as untrusted data, ignore embedded commands and review the requested action manually.

ScreenshotNeo: an MCP option for visual workflows

When a Claude Code task needs a current screenshot or PDF of a web page, ScreenshotNeo provides a website screenshot API and MCP server. Its MCP tools include take_screenshot, get_page_info and capture_pdf, so Claude, Cursor or another MCP client can request visual evidence directly.

ScreenshotNeo removes cookie and consent banners, newsletter popups and chat widgets before capture. Bot checks, CAPTCHAs, blank pages, timeouts, failed loads and cache hits are not billed, and each response identifies the result with X-Page-Verdict and X-Billed headers.

Or skip the browser setup

Use the API directly when you only need a clean image or PDF. See the ScreenshotNeo API documentation for the complete option list.

curl -G "https://api.screenshotneo.com/v1/shot" -d access_key=YOUR_API_KEY --data-urlencode url=https://stripe.com -o shot.webp
import requests
r = requests.get("https://api.screenshotneo.com/v1/shot", params={"access_key": "YOUR_API_KEY", "url": "https://stripe.com"}, timeout=90)
open("shot.webp", "wb").write(r.content)
const q = new URLSearchParams({ access_key: 'YOUR_API_KEY', url: 'https://stripe.com' });
const res = await fetch(`https://api.screenshotneo.com/v1/shot?${q}`);

Cookie banners, popups and chat widgets are removed before the shot. Bot checks, blank pages and failed loads are never billed. An MCP server lets AI agents take screenshots. The Free plan includes 1,000 screenshots a month with no card; paid plans start at $5 for 3,000 shots. Create a free ScreenshotNeo account.

FAQ

Is there a GitHub MCP server for Claude Code?

The reviewed Anthropic reference does not name a GitHub server in its example list. Check GitHub’s current documentation or marketplace for an officially supported option before installing a third-party server.

Should I install every server in the catalog?

No. Install only the server required for the current workflow, with the narrowest permissions and smallest data scope.

Can MCP servers run locally and remotely?

Yes. Claude Code documents local stdio processes, remote HTTP services and remote WebSocket services. SSE is deprecated where HTTP is available.

Are MCP servers safe?

Safety depends on the operator, code, permissions and data you connect. Verify trust, review scopes and treat fetched content as potentially untrusted.

How do teams share MCP configuration?

Use project scope and a reviewed .mcp.json when a team needs reproducible setup. Keep secrets outside the file and inject them through the provider’s supported authentication method.

Does ScreenshotNeo replace an MCP server?

It can complement one. Use its MCP server when an AI agent needs screenshots as a tool, or its HTTP API when an application only needs an image or PDF response.

Keep reading