MCP Reference Servers
Official reference MCP servers: filesystem, git, fetch, memory, and more — the fastest way to give agents real-world tool access.
MCP Reference Servers
The official collection of Model Context Protocol (MCP) reference servers maintained by Anthropic and the MCP community. Drop any of these servers into your agent stack to instantly grant it structured, safe access to local files, git repositories, the web, long-term memory, and databases.
Key features
- Filesystem server — read/write files and directories with path-scoped permissions
- Git server — inspect commits, diffs, branches, and logs
- Fetch server — make HTTP requests from inside a sandboxed tool call
- Memory server — persistent key-value store across agent sessions
- SQLite, Postgres, and more — query databases via natural-language tool calls
- All servers implement the open MCP spec, so they work with Claude, Cursor, Zed, and any MCP client
Quick start
# Install via ai-supply
npx ai-supply add mcp-reference-servers
# Or run a specific server directly with npx (no install required)
npx @modelcontextprotocol/server-filesystem /path/to/allowed/dir
npx @modelcontextprotocol/server-git --repository /path/to/repo
npx @modelcontextprotocol/server-fetch
Add the servers to your Claude Desktop config:
{
"mcpServers": {
"filesystem": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-filesystem", "/Users/me/projects"]
},
"git": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-git", "--repository", "/Users/me/projects/myrepo"]
}
}
}
Curated mirror of the open-source MCP Reference Servers project (MIT/Apache-2.0). Install upstream from the repository.
Compromise signals — malicious or tampered code (leaked secrets, backdoors, a dropped executable) — reduce the score, and known dependency CVEs carry a bounded penalty (they warrant review but never QUARANTINE — update the dependency to clear). Other dangerous-by-capability traits are risk surface, expected for some capabilities. Every finding is mapped to its OWASP control below.
Findings mapped to the OWASP Top 10 for LLM Applications (2025) and the OWASP Machine Learning Security Top 10. Expand any flagged control for the exact findings — compromise reduces the score; expected/risk-surface do not, except a known CVE, which carries a small bounded penalty (high/critical → Review).
The same gate an agent runs before installing (POST /api/v1/trust/mcp-reference-servers/check). Click a policy:
Consume MCP Reference Servers programmatically. Authenticate with an API key or session — see Authorize an agent.
# Agents: CHECK BEFORE YOU INSTALL (no auth) — score, grade, level, capability manifest
curl https://ai-supply.store/api/v1/trust/mcp-reference-servers
# Gate against your org policy (returns { pass, violations })
curl -X POST https://ai-supply.store/api/v1/trust/mcp-reference-servers/check \
-H "Content-Type: application/json" \
-d '{"minGrade":"B","denyPermissions":["shell"],"denyUnknownEgress":true}'
# CLI
npx ai-supply add mcp-reference-servers
# REST (install → download)
curl -X POST https://ai-supply.store/api/v1/listings/mcp-reference-servers/install \
-H "Authorization: Bearer $AIM_KEY"
# MCP tool
install_listing({ "slug": "mcp-reference-servers" })OpenAPI spec →Curated mirror — latest upstream source. See the repository for tagged releases.