Editor's pick
GitHub MCP Server
The most-used developer MCP. Read repos, browse issues and PRs, create PRs, add comments. OAuth setup once, works across all your GitHub orgs.
Category · 12 servers
Development Tools MCP Servers — GitHub, GitLab, Linear, Jira (2026)
Every MCP server in our directory that plugs an AI assistant into your engineering stack — code hosts (GitHub, GitLab, Bitbucket, Azure DevOps), issue trackers (Linear, Jira), observability (Sentry, Datadog), on-call (PagerDuty), CI/CD (CircleCI). Each server has been installed against real production accounts and documented for Claude Desktop, Claude Code, Cursor, and Cline. Provider-agnostic.
12
Servers
4
Sub-types
4
Clients
100%
Human-tested
1What this category covers
Development Tools MCP servers connect AI assistants to the tools engineers use daily — code hosts, issue trackers, observability platforms, on-call systems, CI/CD platforms. They expose read and (with permission) write operations: browse code, read issues, query error rates, acknowledge pages, trigger builds. The MCP server handles auth (usually OAuth or a personal access token) and translates the AI's requests into the tool's API.
The value the MCP layer adds is coherent multi-tool workflows. Individual tool integrations exist as standalone products (Copilot for GitHub, Linear's built-in AI, Datadog's Watchdog). What MCP enables is combining them in one AI conversation: read a Sentry error, find the related code on GitHub, check who's on-call in PagerDuty, create a Linear issue with all the context, comment on the Slack thread — end-to-end in one flow. This cross-tool composition is why MCP matters for engineering.
What lives here vs. elsewhere: if the server's primary users are engineers working on software, it belongs here. If it's about file storage generally (not code-specific), see Files & Storage. If it's about team communication broadly, see Communication. Servers with dual purpose (GitHub for code + docs) appear in the more specific category.
2When you should reach for a Development Tools MCP server
- You want an AI assistant that can read your code, browse your issues, and act on your CI/CD pipelines without you jumping between tools.
- You're building agent workflows that respond to real signals (a Sentry alert, a PagerDuty page, a new issue) with structured actions.
- You want AI-drafted PR descriptions that pull context from linked issues automatically.
- You're doing incident response and want AI to orchestrate the war-room work: pull error data, find responders, draft status updates.
- You need engineering workflows that span multiple tools — MCP is uniquely good at composing them in one AI conversation.
3Featured servers this month
Our editorial picks — heaviest tested, highest reader ratings.
Most requested
Linear MCP Server
The reference for issue tracking. Query pipelines, create issues with proper labels and assignees, transition states. Well-designed API that MCP exposes cleanly.
Highest reliability
GitLab MCP Server
GitHub counterpart for teams on GitLab.com or self-hosted GitLab. Full parity: repos, merge requests, issues, pipelines. Supports enterprise auth patterns.
4All 12 Development Tools servers
Filter by sub-type, search by name, sort by recency or alphabetical.
Code HostsAug 16
GitHub MCP Server
Read repos, browse issues and PRs, create PRs, add comments. OAuth setup once, works across all orgs.
Code HostsAug 12
GitLab MCP Server
GitLab.com and self-hosted GitLab. Repos, merge requests, issues, pipelines. Enterprise auth support.
Code HostsAug 9
Bitbucket MCP Server
Bitbucket Cloud and Data Center. Repos, pull requests, pipelines, workspace-scoped access.
Code HostsAug 7
Azure DevOps MCP Server
Azure Repos, Work Items, Pipelines, Boards. Enterprise SSO support, org-scoped tokens.
Issue TrackingAug 14
Linear MCP Server
Query pipelines, create issues with labels and assignees, transition states. Clean MCP-native design.
Issue TrackingAug 11
Jira MCP Server
Jira Cloud and Data Center. Issues, epics, sprints, JQL queries, custom fields. Enterprise-ready.
Issue TrackingAug 6
Shortcut MCP Server
Shortcut (formerly Clubhouse). Stories, iterations, workflow states. Popular with mid-size product teams.
ObservabilityAug 13
Sentry MCP Server
Query issues, browse traces, check error rates. Environment-aware queries, release comparison support.
ObservabilityAug 10
Datadog MCP Server
Metrics, logs, traces, monitors. Query dashboards, list active alerts, silence noisy monitors.
ObservabilityAug 8
Grafana MCP Server
Grafana Cloud and self-hosted. Query dashboards, browse panels, list active alerts.
CI/CD & On-CallAug 12
PagerDuty MCP Server
Check on-call schedules, acknowledge incidents, browse recent incidents. Read-only by default.
CI/CD & On-CallAug 5
CircleCI MCP Server
Browse pipelines, trigger workflows, inspect job outputs. Context-scoped auth per project.
5Related Skills
AI Skills our team pairs with Development Tools MCP servers for full workflows.
Skill
React Component Generator
Pairs with GitHub MCP to read existing components before generating new ones that match conventions.
Skill
Test Generator with Coverage Analysis
Works with any repo MCP for coverage-aware test generation across the whole codebase.
Skill
PR Summary Generator from Diff
Reads PR context from GitHub/GitLab/Bitbucket MCP, generates PR descriptions with issue links.
Skill
Stack Trace Analyzer with Fix Suggestion
Pulls trace context from Sentry MCP and code from repo MCP for root-cause analysis.
Skill
Incident Response Orchestrator
Cross-tool workflow spanning PagerDuty, Sentry, Slack, and issue tracker MCPs.
Skill
Meeting to Tasks Cross-Tool
Extract action items and create issues in Linear/Jira/Shortcut based on ownership.
6Related Learn guides
Longer-form tutorials for the concepts behind these servers.
Learn guide
OAuth Setup for GitHub MCP Server
The specific OAuth app setup that works across personal accounts and organization repos.
Learn guide
Combining Sentry + GitHub + PagerDuty MCPs for Incident Response
The three-server pattern our team uses for AI-augmented incident response.
Learn guide
Linear vs Jira vs Shortcut for AI Workflows
Which issue tracker's API is friendliest to AI-driven workflows in 2026?
7Comparisons worth reading
Head-to-head reviews that touch on Development Tools workflows.
Comparison
Claude Code vs Cursor for daily coding
Both use the same GitHub MCP — comparison focuses on the client, not the server.
Comparison
GitHub MCP vs Copilot Workspace
MCP-based GitHub access vs Microsoft's dedicated Copilot Workspace tool.
Comparison
Sentry MCP vs Datadog MCP for AI observability
For AI-assisted debugging: which platform's MCP surfaces the most useful context?
8Frequently asked questions
Yes with explicit permission — but every server in this category defaults to read-only for safety. To enable writes (creating PRs, adding comments, triggering pipelines), you explicitly grant those scopes when setting up OAuth or the personal access token. We strongly recommend using separate tokens for read-only vs write-enabled contexts, and never sharing write tokens across environments.
Copilot Workspace and Cursor's native integrations are tightly integrated with specific tools and specific AI providers. MCP servers give you provider-agnostic, tool-agnostic composition — use any AI provider, combine with any other MCP server. For teams already committed to Copilot or Cursor, keep those tools; for teams that want to compose multi-tool workflows or use multiple AI providers, MCP is the better foundation. See our Claude Code vs Cursor comparison for detail.
Yes — each underlying tool has its own rate limits (GitHub: 5,000 req/hour for authenticated, Linear: 1,500 req/hour, Sentry: 40 req/min on free tier). The MCP servers pass through these limits transparently; if you hit them, the AI will get a rate-limit error and typically back off. For high-volume workflows, use paid tiers of the underlying tools. Most interactive workflows stay well under the limits.
For multi-tool orchestration, Claude Opus 4.7 is the strongest choice — extended thinking helps with planning across many tool calls, and Claude stays coherent across long chains better than most. GPT-5 with o5 mode is very competitive. For simple single-tool workflows (just read from GitHub, just check Sentry), any modern provider works. See individual server pages for per-server provider notes.
Register each server with a distinct name in your MCP client's config, then the AI can use tools from any of them in the same conversation. Example: register
github, linear, and sentry; ask the AI to "read this Sentry error, find the related code on GitHub, create a Linear issue with a summary" and it will orchestrate across the three. Our Incident Response Orchestrator Skill is the reference example.Hit an auth or connectivity error with a Development Tools MCP?
Our sister sites cover errors, workarounds, and diagnostic recipes across the AI ecosystem. Bookmark them for the inevitable moment something breaks.