- Home
- Top MCPs for Developer Tools
Best Developer Tools MCP Servers in 2026
Find the best developer-tool MCPs for Claude, Cursor, and VS Code. Ranked for coding workflows, IDE integration, and daily developer productivity — verified for 2026.
About Developer Tools MCP servers
The best MCP servers for developer tools give an AI coding assistant the same surface area a senior engineer has at their terminal: current library documentation, version control, cross-repo code search, dependency vetting, and error triage. The category exists because LLMs hallucinate API signatures, get fooled by stale training data, and waste tokens re-deriving syntax the docs already encode. A good developer-tools MCP closes those gaps with a single config block: Context7 fixes the stale-docs problem, Git and GitHub cover the version-control loop, Semgrep and Socket catch the insecure code and suspicious packages an unsupervised agent will eventually produce, and Sentry turns production errors into something queryable.
Choose by the gap you feel most, then by auth friction. If your model keeps inventing function names that do not exist, install Context7 first — no API key, stdio or remote streamable-http, a three-minute setup. For version control, the Git MCP drives the local binary with no credentials, and the platform layer follows your host: GitHub with a fine-grained personal access token (stdio or remote), GitLab with an api-key for SaaS or self-hosted instances, Bitbucket over OAuth 2.1 for Atlassian shops, Gitea for self-hosted boxes. Sourcegraph adds cross-repo find-references and impact analysis but is remote-only over streamable-http and needs an API key. On the risk side, Semgrep runs static analysis locally over stdio with no account, Socket scores npm and PyPI dependencies for supply-chain risk before the agent installs them, Sentry (OAuth 2.1) makes error triage conversational, and Redis lets the agent inspect keys and queue depths — point it at a development instance, never production. Building a Plaid integration? Its official MCP is sandbox-first: mock financial data and doc search, not live bank connections.
Common mistakes: connecting a developer-tools MCP with credentials scoped to production instead of a sandbox (a Redis MCP aimed at a live cache is one broad key scan away from a latency incident), installing five overlapping docs and search servers until the model cannot pick a tool, and letting the agent install whatever package name it hallucinates — the exact typosquat pattern Socket exists to catch. Every server below lists a one-line claude mcp add command or JSON snippet plus the smoke test that proves it is wired up. Start with one MCP per concrete workflow problem — Context7 if you only pick one — verify it in a single conversation, then layer on the next.
All Developer Tools MCPs
12 MCPs ranked by popularity. Filter by attribute or search by name.
12 of 12 MCPs
| # | MCP | Tags | Setup | Complexity | Labels | |
|---|---|---|---|---|---|---|
| 1 | Git Local Git operations: commit, diff, log, branch, and more. | git, local | 2 min | Low | Official | |
| 2 | GitHub Full GitHub API access: repos, PRs, issues, and code search. | git, github | 5 min | Low | Official | |
| 3 | Context7 Up-to-date library docs pulled directly into your AI context. | coding, documentation | 3 min | Low | ||
| 4 | Redis Inspect keys, run commands, and debug Redis from an AI agent. | database, cache | 3 min | Low | Official | |
| 5 | Bitbucket Manage Bitbucket Cloud repos, pipelines, and pull requests from an AI agent. | git, bitbucket | 10 min | Medium | Official | |
| 6 | GitLab Full GitLab API access: repos, MRs, issues, pipelines, and registries. | git, gitlab | 5 min | Low | ||
| 7 | Sentry Triage errors, inspect traces, and query events from Sentry. | sentry, errors | 3 min | Low | Official | |
| 8 | Socket Score npm, PyPI, Go, and Maven dependencies for supply-chain and vulnerability risk before an agent installs them. | socket, security | 3 min | Low | Official | |
| 9 | Plaid Official Plaid MCP — developer sandbox tooling and dashboard diagnostics, not a live bank-data connector. | plaid, fintech | 10 min | Medium | Official | |
| 10 | Semgrep Open-source SAST scanning against custom rules from an AI agent. | security, sast | 5 min | Low | Official | |
| 11 | Gitea Self-hosted Git server — manage repos, issues, and PRs from an AI agent. | git, gitea | 10 min | Medium | ||
| 12 | Sourcegraph Cross-repo code search, navigation, and Deep Search via the official Sourcegraph MCP. | sourcegraph, code-search | 10 min | Medium | Official |
Choose the right MCP
Quick decision guide based on your use case.
| If you need… | Start with |
|---|---|
| You need accurate library docs | Start with Context7 |
| You want safe code execution | Use E2B |
| You want to stage commits from the agent | Use the Git MCP |
| Your code lives on GitLab, Bitbucket, or self-hosted Gitea | Use the matching platform MCP — GitLab, Bitbucket, or Gitea |
| You want dependencies vetted before the agent installs them | Use Socket |
| You want static analysis before every commit | Use Semgrep |
| You are on call and triaging production errors | Use Sentry |
| You need find-references across many repos | Use Sourcegraph |
Top Developer Tools MCPs ranked
Detailed cards with setup time, complexity, and key labels.
Up-to-date library docs pulled directly into your AI context.
Inspect keys, run commands, and debug Redis from an AI agent.
Manage Bitbucket Cloud repos, pipelines, and pull requests from an AI agent.
Full GitLab API access: repos, MRs, issues, pipelines, and registries.
Triage errors, inspect traces, and query events from Sentry.
Score npm, PyPI, Go, and Maven dependencies for supply-chain and vulnerability risk before an agent installs them.
Official Plaid MCP — developer sandbox tooling and dashboard diagnostics, not a live bank-data connector.
Open-source SAST scanning against custom rules from an AI agent.
Self-hosted Git server — manage repos, issues, and PRs from an AI agent.
Cross-repo code search, navigation, and Deep Search via the official Sourcegraph MCP.
Archived (historical reference)
2 Developer Tools entries are archived — the upstream package was deprecated or pulled, or a documented security issue applies. The detail page is preserved for historical reference and migration guidance, but these are NOT current editorial picks.
Also in the ChatGPT Apps directory
1 Developer Tools brand is available as one-click ChatGPT Apps. These are not ranked alongside the editorial picks above — they're listings from chatgpt.com/apps surfaced here for brand-search continuity.
FAQ: Developer Tools MCPs
Which MCP should I install first as a developer?
Context7 for accurate library docs, Filesystem for local code access, and GitHub or Git for version control. Those three cover 80% of daily developer workflows.
Do I need a developer-focused MCP if I use Cursor?
Yes. Cursor handles the editor UX; MCPs extend what the agent inside Cursor can actually do — fetching current docs (Context7), running code safely (E2B), or touching your filesystem beyond the open project.
Is it safe to let an agent install npm packages it suggests?
Not unsupervised. Models occasionally suggest package names that do not exist, and typosquatted versions of those names sometimes do. Put Socket in the loop — it scores npm, PyPI, Go, and Maven packages for supply-chain and malware risk before install — and keep a human confirmation on anything with install scripts or low download counts.
Can an AI agent help with on-call and production errors?
Yes. The Sentry MCP (OAuth 2.1, stdio or remote streamable-http) lets the agent query issues, traces, and error trends, which turns the what-broke-overnight question into a two-minute conversation. Keep it read-mostly at first: let the agent summarize and propose, and leave resolving or assigning issues to a human.
Do I need Sourcegraph if I already have the GitHub MCP?
They answer different questions. GitHub code search finds text; Sourcegraph resolves references — who calls this function, where a symbol is defined across dozens of repos — which is what you want before a breaking change. It is remote-only over streamable-http with an api-key, and earns its slot in multi-repo orgs rather than single-repo side projects.
