Editorial MCP shortlist

MCP servers for Claude Code

Connect the systems that explain your coding task: the issue that defines it, the error that reproduces it, the browser flow that fails or the deployment that needs investigation. This collection starts with those workflows and the connection each requires.

For Claude Code users who need external context or actions alongside their existing development workflow.

Lead with the workload

The decision to make

Begin with one connector and a testable read task. A server configuration being saved does not prove that authentication succeeded or that the expected tools are available. Check the connection and tool list before using it in a larger workflow.

Editorial sources reviewed . This selection is based on documentation review, not comparative benchmark testing.

Connect a remote service over HTTP

For a Linear investigation that only needs to read issues, its documented read-only endpoint is a focused starting point. Add it to Claude Code, then open /mcp to complete authentication and inspect the connection. This command configures the client; it does not create or update a Linear issue.

claude mcp add --transport http linear-readonly https://mcp.linear.app/mcp/readonly

Launch a local server over stdio

Playwright MCP runs locally and is launched by the client. Install the runtime required by its documentation, then add the server. Keep the package command after the separator so its arguments are passed to the server process.

claude mcp add --transport stdio playwright -- npx -y @playwright/mcp@latest

Check one complete round trip

Use /mcp to inspect status, authentication and the tools exposed by the connector. Ask for one known issue or a browser snapshot of a test page, and check the result. If the server fails, fix the endpoint, credentials or local runtime before adding further integrations.

Keep this selection distinct from older GitHub examples

The GitHub reference profile in this directory describes an archived implementation. GitHub’s maintained github/github-mcp-server is a different project. Follow the provider’s current documentation if GitHub access is the capability you need; do not install the archived example from an old tutorial.

Selected implementations

6 connectors by workflow

Full MCP hub →
Connect planning to current issues

MCP Linear Server

Official Maintainer: Linear

Start here when a coding task begins in Linear. Read the issue and project context before changing code, then separately review any proposed tracker update.

Before connecting: Use the read-only endpoint when the task is research. Enable writes only for a workflow that actually creates or updates Linear records.

Read the profile and setup notes → · Maintainer documentation ↗

Investigate application errors

MCP Sentry Server

Official Maintainer: Sentry

Add production error context to a debugging session: identify a Sentry issue, inspect the relevant evidence and connect it to the working tree.

Before connecting: Confirm access to the intended organization and project. Hosted and self-hosted Sentry configurations do not expose identical features.

Read the profile and setup notes → · Maintainer documentation ↗

Explore and debug browser workflows

MCP Playwright Server

Official Maintainer: Microsoft / Playwright

Use this when a coding task needs interactive browser investigation. Inspect the controls, reproduce the journey and keep the browser session available between steps.

Before connecting: Use a separate browser profile for the task. For repeatable automated tests, compare the MCP workflow with ordinary Playwright scripts and the project’s CLI guidance.

Read the profile and setup notes → · Maintainer documentation ↗

Understand a deployment failure

MCP Vercel Server

Official Maintainer: Vercel

Useful after a deployment fails: inspect the relevant deployment and logs without reducing the investigation to a pasted build-error excerpt.

Before connecting: Complete OAuth with the intended account and check supported clients. The hosted endpoint replaces old, unverified local-package instructions.

Read the profile and setup notes → · Maintainer documentation ↗

Join Jira work with Confluence context

MCP Jira Server

Community Maintainer: MCP Atlassian contributors

Choose this community implementation when the coding brief spans Jira issues and Confluence pages, including supported Server/Data Center environments.

Before connecting: Choose the correct authentication method for Cloud or Server/Data Center. This implementation is not an Atlassian-maintained product.

Read the profile and setup notes → · Maintainer documentation ↗

Research public pages and extract their content

Firecrawl MCP Server

Official Maintainer: Firecrawl

Useful for a bounded documentation-research task across several public pages. Retrieve the source material before asking the coding agent to synthesize a change.

Before connecting: Choose the appropriate hosted or local configuration. Bound page counts and review credit usage; a website crawl and a single page fetch have different costs.

Read the profile and setup notes → · Maintainer documentation ↗

Archived implementations: migration notes

These historical profiles are excluded from the shortlist. Do not select an archived reference implementation for a new production deployment. Check the current provider or project documentation when replacing one.

Sources and next steps

Each connector above links to its maintainer’s own documentation. Continue with the MCP server hub or the MCP explainer.