MCP servers that make coding agents better
MCP servers that extend Claude Code, Cursor, and Copilot: docs lookup, code search, sandboxed execution, Git, and language server tools for agents.
What to know about Coding agent MCP servers
Coding agents already read and edit files, so the MCP servers worth adding are the ones that give them context they lack. Documentation servers fetch current library references, code search servers index large or multi-repository codebases, and language server bridges expose go-to-definition, references, and diagnostics so the agent navigates code like an IDE does.
A second group extends what the agent can do: sandboxed code execution for running snippets safely, Git hosting servers for issues and reviews, and servers for package registries that report versions and vulnerabilities. Each removes a round trip where you would otherwise copy information into the chat by hand.
Resist installing everything. Every server adds tools to the context, and coding agents with dozens of overlapping tools pick the wrong one more often. Start with a docs server and one code intelligence server, and add others only when you notice a gap.
Coding agent MCP servers
The 12 strongest of 1557 matching servers. Servers that name the topic directly come first, then by trust score.
- 65Trust score 65 of 100, Good
Local memory for coding agents; a rule with a proof can refuse. Full install with hooks: the repo.
RegistryLicenseQualityMaintenanceLicense: C. GPL-3.0 is a strong copyleft license: derived works must use the same license.Quality: A. 91/100 for listing completeness. Lost points: short summary.Maintenance: A. Last commit or release 12 days ago. - 63Trust score 63 of 100, Good
Windows Sandbox MCP
by fmuecke · C#
Run risky commands in a disposable, offline Windows Sandbox (Windows only, needs a one-time setup).
RegistryLicenseQualityMaintenanceLicense: A. MIT is a permissive open source license.Quality: Not graded. No README or tool list has been captured yet.Maintenance: A. Last commit or release 5 days ago. - 61Trust score 61 of 100, Good
Delegate small coding tasks to local Ollama/llama.cpp models; patches gated by real test runs.
RegistryLicenseQualityMaintenanceLicense: A. MIT License is a permissive open source license.Quality: Not graded. No README or tool list has been captured yet.Maintenance: Not graded. No commit or release date has been recorded. - 60Trust score 60 of 100, Good
Search the Claude Code Ultimate Guide and machine-readable references from any MCP client.
RegistryLicenseQualityMaintenanceLicense: Not graded. No license is recorded for this server.Quality: Not graded. No README or tool list has been captured yet.Maintenance: Not graded. No commit or release date has been recorded. - 55Trust score 55 of 100, Fair
Budget-capped agent wallet: pay x402 services in tANE, a free testnet token with no cash value
RegistryLicenseQualityMaintenanceLicense: Not graded. No license is recorded for this server.Quality: Not graded. No README or tool list has been captured yet.Maintenance: Not graded. No commit or release date has been recorded. - 55Trust score 55 of 100, Fair
Run parallel, resumable, human-operable Antigravity CLI sessions from any MCP agent harness
RegistryLicenseQualityMaintenanceLicense: Not graded. No license is recorded for this server.Quality: Not graded. No README or tool list has been captured yet.Maintenance: Not graded. No commit or release date has been recorded. - 55Trust score 55 of 100, Fair
Cited, searchable memory of a codebase: hybrid search, git history, impact analysis, doc drift.
RegistryLicenseQualityMaintenanceLicense: Not graded. No license is recorded for this server.Quality: Not graded. No README or tool list has been captured yet.Maintenance: Not graded. No commit or release date has been recorded. - 55Trust score 55 of 100, Fair
Audits Claude Code/Cursor MCP configs for supply-chain backdoors and risky permissions.
RegistryLicenseQualityMaintenanceLicense: Not graded. No license is recorded for this server.Quality: C. 60/100 for listing completeness. Lost points: no listed tools, short summary, no docs link.Maintenance: Not graded. No commit or release date has been recorded. - 55Trust score 55 of 100, Fair
Self-hosted semantic code search MCP for Claude Code & Cursor. Multi-repo, AST-aware, stays local.
RegistryLicenseQualityMaintenanceLicense: Not graded. No license is recorded for this server.Quality: Not graded. No README or tool list has been captured yet.Maintenance: Not graded. No commit or release date has been recorded. - 55Trust score 55 of 100, Fair
Remote Linux boxes for coding agents: Docker, a browser, screenshots, logs, human takeover.
RegistryRemoteLicenseQualityMaintenanceLicense: Not graded. No license is recorded for this server.Quality: Not graded. No README or tool list has been captured yet.Maintenance: Not graded. No commit or release date has been recorded. - 55Trust score 55 of 100, Fair
Create and update roadmap tree nodes and statuses in the RoadRaven desktop app (must be running).
RegistryLicenseQualityMaintenanceLicense: Not graded. No license is recorded for this server.Quality: Not graded. No README or tool list has been captured yet.Maintenance: Not graded. No commit or release date has been recorded. - 55Trust score 55 of 100, Fair
Mock small-business CRM, inbox, calendar, invoicing and phone tools for testing AI agents safely.
RegistryLicenseQualityMaintenanceLicense: Not graded. No license is recorded for this server.Quality: Not graded. No README or tool list has been captured yet.Maintenance: Not graded. No commit or release date has been recorded.
How to install one
- Compare the shortlist. THOR leads this topic. Open two or three listings and compare their tools, transport, and required API keys.
- Copy the config for your client. Each listing has an install guide, for example Claude Desktop, Claude Code, Cursor, or VS Code. The client guides show where each config file lives.
- Scope credentials, then restart. Use the narrowest token or folder the server needs, restart the client, and check that the new tools appear.
Coding agent MCP questions
Which MCP servers are most useful in Claude Code?
Documentation lookup, a code intelligence or language server bridge, and a Git hosting server cover most gaps. The Claude Code client page shows how to add each one from the command line.
Why would a coding agent need a sandbox server?
A sandbox runs untrusted or experimental code in an isolated container rather than on your machine. It is useful for testing snippets, data scripts, and dependencies you have not vetted.
Can too many MCP servers slow a coding agent down?
Yes. Each server's tool definitions take context space, and overlapping tools make selection harder. Most clients let you disable servers per project, which keeps the active set small.