Software development is the birthplace of MCP and by far its most mature category. GitHub, Sentry, Linear, Atlassian and Figma maintain official servers, and virtually every database has one. For development teams this is the place to start: quality is high and the use cases are proven.
Atlassian Rovo MCP Server
Official
by Atlassian
Official remote MCP server (https://mcp.atlassian.com/v1/mcp) connecting Jira, Confluence, Jira Service Management, Bitbucket and Compass to AI tools while respecting user permissions.
Capabilities: search/summarize Jira issues · create/update Jira issues · search/create/update Confluence pages · Compass components · Bitbucket and JSM access
- Authentication
- OAuth 2.1 or API tokens
- Use cases
- sprint and backlog management from AI clients, documentation summarization, cross-tool search
⚠ GA since February 2026. For Server/Data Center use community sooperset/mcp-atlassian instead.
Documentation / source →
Chrome DevTools MCP
Official
by Google (Chrome DevTools team)
Official Chrome DevTools MCP server that lets coding agents control and inspect a live Chrome browser: performance traces, network requests, console and screenshots.
Capabilities: performance tracing and insight analysis · list network requests · read console messages · screenshots · Puppeteer-based automation
- Authentication
- None (local, npm chrome-devtools-mcp)
- Use cases
- frontend performance debugging, runtime debugging of web apps by AI agents
⚠ Public preview since Sept 2025; officially supports only Google Chrome / Chrome for Testing.
Documentation / source →
CircleCI MCP Server
Official
by CircleCI
Official CircleCI MCP server giving AI tools direct access to build logs, job metadata and failure context for CI debugging.
Capabilities: retrieve build logs · find failed pipelines · job metadata and failure context
- Authentication
- CircleCI API token (local stdio server, npm)
- Use cases
- fix failed builds from Cursor/Claude Code/Windsurf, CI pipeline debugging without UI
Documentation / source →
Figma MCP Server (Dev Mode)
Official
by Figma
Figma's official MCP server (remote at https://mcp.figma.com/mcp or local via desktop app) that gives coding agents design context and can write native content back to the canvas.
Capabilities: extract design context for design-to-code · read frames/components/variables/auto layout · write-to-canvas: create and modify Figma content (remote only)
- Authentication
- Personal access token (file:read scope) / OAuth; works on all plans incl. free Starter
- Use cases
- design-to-code in VS Code/Cursor/Claude Code, keeping code consistent with design system
⚠ Write-to-canvas is beta (free during beta). Official Claude Code plugin available.
Documentation / source →
GitHub MCP Server
Official
by GitHub (Microsoft)
GitHub's official MCP server (remote at https://api.githubcopilot.com/mcp/ or local Docker) connecting AI agents to repositories, issues, pull requests and GitHub Actions.
Capabilities: read/search repositories and code · create and manage issues and pull requests · monitor GitHub Actions workflow runs and build failures · manage releases · modular toolsets
- Authentication
- OAuth 2.0 (remote) or Personal Access Token
- Use cases
- issue triage, PR review automation, CI failure debugging from Claude/Cursor/Copilot
⚠ Remote hosted server is the recommended path; separate enterprise configuration docs.
Documentation / source →
GitLab MCP Server
Official
by GitLab
GitLab's built-in MCP server (part of Duo Agent Platform) that lets AI tools securely access project data, issues and merge requests on a GitLab instance.
Capabilities: retrieve project information · read/update issues and merge requests · secure GitLab API access · AI Catalog MCP servers · glab CLI mcp command
- Authentication
- OAuth 2.0 with Dynamic Client Registration (HTTP transport recommended)
- Use cases
- issue and MR management from AI clients, DevSecOps workflow automation
⚠ Beta since GitLab 18.6; MCP spec 2025-03-26/2025-06-18 supported since 18.7. Community alternative: zereight/gitlab-mcp.
Documentation / source →
by Linear
Linear's hosted MCP server (https://mcp.linear.app/mcp, Streamable HTTP + SSE) for finding, creating and updating issues, projects and comments.
Capabilities: find/create/update issues · manage projects · add comments
- Authentication
- OAuth interactive flow, or Bearer header with OAuth token/API key
- Use cases
- issue management from Claude Code, Cursor, Windsurf, Zed, v0
⚠ Older community servers (tacticlaunch/mcp-linear, jerhadf/linear-mcp-server) are superseded.
Documentation / source →
MongoDB MCP Server
Official
by MongoDB
Official MongoDB MCP server to query and manage data and deployments on Atlas, Community Edition and Enterprise Advanced from AI clients.
Capabilities: query and update data in natural language · Atlas cluster management · database administration
- Authentication
- MongoDB connection string; Atlas API service-account credentials for Atlas management
- Use cases
- data exploration from AI clients, Atlas ops automation
⚠ Read-only flag available; works with all deployment types.
Documentation / source →
by Neon
Official MCP server for Neon serverless Postgres: manage projects and branches, run SQL and migrations via the Neon Management API in natural language.
Capabilities: manage Neon projects and branches · run SQL · database migrations · Neon Management API operations
- Authentication
- OAuth (remote, no API key needed) or Neon API key (local)
- Use cases
- agent-driven Postgres provisioning, branch-based database workflows
⚠ Open source; also available on Docker MCP Catalog.
Documentation / source →
by Microsoft
Microsoft's official browser-automation MCP server using structured accessibility snapshots instead of screenshots, so no vision model is needed.
Capabilities: navigate and interact with web pages · form filling and clicks · screenshots · test generation · accessibility-snapshot based control
- Authentication
- None (runs locally, drives a local browser)
- Use cases
- AI-driven E2E testing, web scraping/automation from coding agents, verifying UI changes
⚠ De-facto successor of the archived Anthropic puppeteer reference server.
Documentation / source →
Postman MCP Server
Official
by Postman
Official Postman MCP server (remote at https://mcp.postman.com or npm) giving agents access to workspaces, collections, environments and API evaluation.
Capabilities: manage collections and environments · access workspaces · evaluate APIs · automate API workflows
- Authentication
- Postman API key
- Use cases
- API test automation, collection maintenance via natural language
⚠ Postman also curates a trusted list of third-party official MCP servers (postman.com/getmcp).
Documentation / source →
Redis MCP Server
Official
by Redis
Official Redis MCP server: natural-language interface to manage and search Redis data structures (strings, hashes, lists, sets, sorted sets, streams, JSON, pub/sub).
Capabilities: manage all core Redis data structures · search and retrieval · pub/sub and streams · vector search
- Authentication
- Redis connection credentials; separate mcp-redis-cloud uses Redis Cloud API keys
- Use cases
- cache inspection by agents, agent memory/state management on Redis
⚠ Replaces older community reference server (GongRzhe/REDIS-MCP-Server).
Documentation / source →
by Sentry
Official Sentry MCP server (remote at https://mcp.sentry.dev/mcp) exposing 16+ tools to query issues, errors, projects and teams from LLM clients.
Capabilities: query issues and errors · manage projects and teams · error root-cause analysis (Seer) · organization data access
- Authentication
- OAuth 2.0 only on remote endpoint; stdio with User Auth Token for self-hosted Sentry
- Use cases
- automated error analysis in coding agents, debugging production issues from Cursor/Claude Code
⚠ Remote endpoint does not support API-token auth (open issue #833); runs on Cloudflare Workers.
Documentation / source →
Supabase MCP Server
Official
by Supabase
Official Supabase MCP server for managing projects, database schema, queries, auth, storage and edge functions from AI agents; remote server with OAuth2 available.
Capabilities: create projects and design tables · run SQL queries · manage auth/storage/edge functions · configuration management
- Authentication
- OAuth2 (remote MCP auth) or personal access token (local)
- Use cases
- AI-assisted backend development, database schema iteration from Cursor/Claude
⚠ Write access to your database — use read-only mode and project scoping where possible.
Documentation / source →
by crystaldba (open source)
Leading open-source Postgres MCP server with configurable read/write access, index tuning, explain plans and database health analysis for any PostgreSQL.
Capabilities: configurable read/write SQL access · index tuning · explain plans · database health checks · performance analysis
- Authentication
- Database connection string; restricted/unrestricted modes
- Use cases
- safe agent access to Postgres, performance tuning during development and production
⚠ No single official Postgres MCP server exists; Anthropic's reference postgres server is archived. Alternatives: pgEdge Postgres MCP, AWS Aurora Postgres MCP.
Documentation / source →