Top file management MCP servers

Ranked by the composite AgentRank score — a weighted blend of stars (15%), freshness (25%), issue health (25%), contributors (10%), and inbound dependents (25%). Average score across all 204 file/storage tools is 28.4. The tools below are in the top tier.

Choosing by use case

Google Drive and cloud documents

xing5/mcp-google-sheets leads the file management category at 83.27 with 17 contributors and a commit on March 15. Despite the name, it covers the full Google Drive API — listing, reading, creating, and modifying files. The Sheets-specific tools are a layer on top. If your agent needs to interact with Google Workspace documents, this is the highest-signal option in the index.

Local filesystem (power user)

wonderwhy-er/DesktopCommanderMCP scores 78.48 with 5,721 stars — the most-starred filesystem server in the index. It pairs file operations (search, read, diff-aware editing) with terminal control in a single server. If you need an agent that can navigate a codebase, edit files with minimal diff noise, and run shell commands as part of the same workflow, this is the pick. 31 contributors signals ongoing community investment.

AWS S3 and cloud storage

awslabs/mcp is the official AWS Labs MCP suite at 71.72 with 8,471 stars and 243 contributors. It covers S3 object storage alongside CloudWatch, Lambda, CDK, and other AWS services. Last committed March 18 — the most recently updated server in this list. If your agent workflow touches S3 buckets (reading objects, listing, writing), this is the safest bet given the official backing and contributor depth.

Official reference implementation

modelcontextprotocol/servers is the canonical reference at 70.26 with 81,030 stars and 1,018 contributors. The @modelcontextprotocol/server-filesystem package provides read, write, list, search, and move operations on local directories you explicitly allow. The @modelcontextprotocol/server-gdrive package provides Google Drive search and read. These are the servers most MCP clients assume you've configured — they're in the official docs and quickstarts everywhere.

Pure filesystem (Go)

mark3labs/mcp-filesystem-server scores 55.88 with 616 stars. It's a focused Go implementation with zero external dependencies covering the full file operation set: read, write, create, delete, move, copy, list. No terminal integration, no cloud storage — just file operations done cleanly. Note the last commit was November 2025; worth watching whether maintenance resumes.

High-performance filesystem (Rust)

rust-mcp-stack/rust-mcp-filesystem scores 54.68 with a commit on March 15. The async Rust implementation keeps memory usage low and handles concurrent file operations without blocking. At 138 stars and 3 contributors it's newer, but the freshness score is strong. Best fit for agents doing heavy parallel file I/O where the overhead of a Python or Node implementation would be measurable.

What to watch for

Scope creep vs focus

There's a spectrum here: DesktopCommanderMCP does filesystem + terminal in one server (convenient but broad), while mcp-filesystem-server does only files (focused but narrow). The right choice depends on whether you want your agent to have shell access in the same session. If you're security-conscious, a file-only server with explicit path allowlists is safer.

Path allowlisting

Most filesystem servers require you to configure which directories the agent can access. Do not skip this. Giving an agent access to your entire home directory is a significant risk surface. The official modelcontextprotocol/servers filesystem implementation enforces explicit path declarations — treat this as a baseline expectation for any server you run in production.

Cloud storage vs local filesystem

Cloud storage tools (mcp-google-sheets, awslabs/mcp) require OAuth credentials or IAM roles and introduce network latency. Local filesystem tools are faster but limited to what's on the machine running the MCP server. For most agent workflows, you'll want both — local for in-session file editing, cloud for persistent storage and sharing.