Skills · last verified 2026-10-23
modelcontextprotocol/servers — install for Claude Desktop
github.com/modelcontextprotocol/servers · MCP server collection · reference servers
What it does
The Model Context Protocol reference server collection: example servers maintained alongside the protocol, including filesystem, git, memory, sequential-thinking, time, fetch and everything. This page focuses on the filesystem server, the most directly useful one for a coding agent, because the collection as a whole is too large to review as one item.
The filesystem server reads and writes files, creates, lists and deletes directories, moves files and directories, searches files and reads file metadata — but only inside directories passed to it on the command line. Those allowed directories are the whole access control story: a server started without any answers file tool calls with an error, and the README recommends the command-line form over the deprecated Roots mechanism.
Install
Filesystem server, Claude Desktop config (verified against the repo README's Claude Desktop example):
{
"mcpServers": {
"filesystem": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-filesystem", "/path/to/allowed/files"]
}
}
}
Replace /path/to/allowed/files with the directories the agent is actually allowed to touch, and add more paths as needed. The README gives the same shape for the other servers (for example, the git server runs via uvx with a repository path).
Verified against github.com/modelcontextprotocol/servers (README; src/filesystem/README.md), 2026-10-23.
Trust score: NO_SCORE
No score is given for this page. This is a collection of server codebases, not a set of instruction files, and the filesystem server source was not fully read in this pass, so a scored review would not be honest. What was checked is listed under Verification below. Note the capability itself: read/write file access inside the allowed directories includes deleting and moving files, so the allowed list should be as narrow as the job allows.
Verification
Tier: AUTHOR_DOCS_VERIFIED. Verified 2026-10-23 from the author's own repository (fresh clone this run). No install was executed and no account was connected in this pass.
What was checked: Author README (server list, Claude Desktop configuration examples) and the filesystem server README (features, directory access control) were read from a fresh clone of the repo. The server source code was not fully read, and the other servers in the collection were not individually reviewed.
Limits: Configuration checked against the author README; not executed in this pass. No full code read was done, so no trust score is given and no score is claimed for the other servers in the collection.
Sources: github.com/modelcontextprotocol/servers (README; src/filesystem/README.md)
Also in this series
- wshobson/agents — install for Claude CodeNO_SCORE · 2026-10-23
- vercel-labs/agent-skills — install for Claude CodeNO_SCORE · 2026-10-23
- anthropics/skills — install for Claude CodeNO_SCORE · 2026-10-23
- microsoft/playwright-mcp — install for Claude CodeNO_SCORE · 2026-10-23
- supabase-community/supabase-mcp — install for MCP clientsNO_SCORE · 2026-10-23
- exa-labs/exa-mcp-server — install for Claude CodeTrust 7/10 · 2026-10-23
- getsentry/sentry-mcp — install for Claude CodeNO_SCORE · 2026-10-23
- stripe/agent-toolkit — install for Claude CodeNO_SCORE · 2026-10-23
- github/awesome-copilot — install for GitHub CopilotNO_SCORE · 2026-10-23
- qdrant/mcp-server-qdrant — install for Claude DesktopNO_SCORE · 2026-10-23
- neondatabase/mcp-server-neon — install for Claude CodeNO_SCORE · 2026-10-23
Back to: Skills index