Best MCP Servers for Developers

The picks

Ranked by fit, not by popularity. Each entry links to its full Directories page.

1. 1

   codebase-index

   Symbol-level code search and context across your entire codebase.

   Developers constantly need to answer 'where is function X', 'what calls Y', or 'show me the schema for Z'. CodeGraph-style indexing (pre-indexed symbols, call graphs, cross-reference tables) eliminates the grep tax and returns precise results with minimal latency. This is the foundational tool every developer needs before writing new code. It's faster than reading files sequentially and more precise than keyword search.

2. 2

   git-client

   Git history, diffs, blame, and branch context without touching your terminal.

   Understanding code history is inseparable from understanding code intent. A Git client MCP lets you run git log, git show, git blame, and git diff without shell friction, and Claude can directly parse the output to answer 'why was this line added', 'what changed between branches', or 'who introduced this bug'. Commits and diffs are the source of truth for intent; conversations should have first-class access.

3. 3

   database-client

   Run queries, inspect schemas, and debug data state in seconds.

   Backend developers spend significant time querying databases to understand schema, verify data migrations, or investigate bugs. A database MCP client (supporting Postgres, MySQL, SQLite, etc.) lets you introspect tables, run read-only queries, and see results formatted cleanly without context-switching to a DB admin tool. The dev-time speedup is massive when Claude can directly explore your data alongside your code.

4. 4

   test-runner

   Run unit tests, see failures, and get assertions directly in the conversation.

   When you ask 'why is this test failing', you want the failure output right there, not a separate browser tab. A test-runner MCP executes tests (Jest, Pytest, Go test, etc.) and returns structured results and failure details. Claude can read the actual failure message and suggest fixes immediately. This is faster and more reliable than manually copying test output.

5. 5

   github-api

   PR status, CI checks, issue context, and repository metadata without browser context-switches.

   Developers need to know 'why did CI fail', 'is my PR blocked', 'what are the open issues in this repo'. A GitHub API MCP gives you PRs, checks, branches, issues, and workflows as structured data. You can ask Claude 'why did the build fail' and get the actual CI log without hunting through the GitHub UI. It's especially useful for teams running automated checks.

6. 6

   docker-runtime

   Run containers, inspect logs, and test builds without leaving your LLM conversation.

   Containerized development is the norm; a Docker MCP lets you build, run, and inspect containers directly. You can ask 'does this Dockerfile work' and Claude can test-build it, inspect the image, and see build errors without you manually running docker commands. This is valuable for debugging container issues and verifying configuration.

7. 7

   terminal-executor

   Execute shell commands, compile code, and run scripts with live output streaming.

   Sometimes developers need to run arbitrary commands: compile code, run linters, execute scripts, or check tool versions. A safe terminal executor MCP (with configurable command allowlists) lets Claude run these commands directly and see the output. It bridges the gap between 'I need to run a command' and 'I need to leave my conversation to do it'. Use with caution on untrusted prompts, but essential for local development loops.

Other lists

More curated picks across the index.