← Back to directory
F

Filesystem MCP Server

Community
Tutorial / collection
Local project-file access over stdio.

This is a tutorial or resource collection; there is nothing to install into a client.

Category
Filesystem #23 of 24
Stars
★ 42.7k Very popular
Transport
stdio (local process)
Runtime
Node.js · Python
Credentials
No credential needed
License
Apache-2.0
Last commit
35FMRS · D

The source identifies this as a stdio filesystem configuration example in an Anthropic-managed repository, but does not establish that this specific server is maintained by the upstream filesystem protocol owner; tool names and client compatibility are not evidenced.

Strongest · Maintenance 11/20 Weakest · Documentation 5/20

Reliability
6/20
Security and permissions
6/20
Maintenance
11/20
Documentation
5/20
Setup experience
7/20
Why each score
Reliability 6/20
The manifest only provides three launch configurations; it provides no server source, MCP initialization or tool-list implementation, tests, or CI evidence. The filesystem entry uses an unpinned npx dependency, the database entry depends on unprovided local script and config files, and custom-tools depends on an unverified Python module; its --port argument also creates uncertainty for a stdio configuration. A happy-path launch and handshake are therefore only plausible, with deductions for unverified dependencies, error handling, and actual tool behavior.
Security and permissions 6/20
The evidence shows filesystem access scoped to CLAUDE_PROJECT_DIR, database access through DATABASE_URL, and custom-tools access through API_KEY, with possible external database or service access. Environment variables avoid embedding credentials directly, which is positive, but there is no evidence of finer-grained permissions, path boundaries, network scope, confirmation for dangerous actions, auditing, or data-flow disclosure. No behavior clearly triggers a red-line finding and no real secret is shown, so this is not a red-line score; insufficient control evidence still warrants substantial deductions.
Maintenance 11/20
The repository is described as Anthropic-managed, Apache-2.0 licensed, and not archived, providing ownership and licensing evidence. However, the supplied material contains no release history, commit cadence, dependency-update policy, issue-response evidence, vulnerability-reporting process, or security-response channel. Stars cannot establish maintenance quality, so the score remains moderate.
Documentation 5/20
The README documents the general plugin directory, installation command, and repository structure, but it does not document this manifest's prerequisites, the tools or parameters of its three entries, authentication requirements, permission limits, platform compatibility, troubleshooting, or cost. It also explicitly warns that included MCP servers are not verified, limiting reviewable evidence. The documentation score is therefore low.
Setup experience 7/20
The README gives a generic Claude Code marketplace installation path, but no server-specific client configuration, Node/Python runtime requirements, dependency setup, database configuration, API-key instructions, or stdio connection verification. npx, plugin-root files, and the Python module introduce undocumented environmental assumptions, making the setup inferable but fragile.

Test out of dateListed 2026-07-29

Read the FMRS scoring method →

Fit and risk

What it can accessReads local filesWrites / deletes local filesUses the networkConnects to a database

Best for

  • Developers who need an MCP client to access a specified Claude Code project directory.

Not for

  • Users who need database access or custom API tools.
  • Users who need remote deployment, SSE, or streamable-http transport.

Required permissions

  • Access to the local project directory specified by ${CLAUDE_PROJECT_DIR}.
  • The exact read/write scope depends on the client, operating system, and server implementation; the source does not specify it.

Risks and side effects

  • The MCP server may access local files; specify only a directory you are willing to authorize.
  • The README warns that MCP servers, files, or other software included in plugins may not be controlled by Anthropic and are not guaranteed to work as intended or avoid making changes.

Setup

Before you start

Runtime:Node.js · Python

DATABASE_URL optionalsecret Database connection string supplied by the user's own database, injected via ${DATABASE_URL}.
CUSTOM_API_KEY optionalsecret API key for the custom MCP server, provided by the user's own service; sensitive credential.
Other optional settings (3)
LOG_LEVEL optional Example log level, e.g. info, controlling MCP server log verbosity.
DB_POOL_SIZE optional Database connection pool size, set to 10 in the example.
DEBUG optional Debug toggle, set to false in the example.

Add this configuration to a client that supports stdio MCP servers, and set ${CLAUDE_PROJECT_DIR} to the target project directory. Before running it, confirm that Node.js and npx are installed and that you trust the relevant plugin and server software.

MCP config · JSON
{"command":"npx","args":["-y","@modelcontextprotocol/server-filesystem","${CLAUDE_PROJECT_DIR}"],"env":{"LOG_LEVEL":"info"}}

Works in clients that read an "mcpServers" key. For clients such as VS Code that use a different key, the configurator below converts it.

.vscode/mcp.json
{
  "command": "npx",
  "args": [
    "-y",
    "@modelcontextprotocol/server-filesystem",
    "${CLAUDE_PROJECT_DIR}"
  ],
  "env": {
    "LOG_LEVEL": "info"
  }
}

Goes in your project's .vscode/mcp.json (VS Code uses a "servers" key).

Check that it works

After configuring the filesystem MCP server in Claude Code, check that filesystem tools appear in the client's tool list, then ask it to list the current directory files to confirm the connection works.

Troubleshooting

  1. Confirm that ${CLAUDE_PROJECT_DIR} resolves to a valid directory.
  2. Confirm that Node.js and npx are installed.
  3. Check that the client supports stdio MCP configuration.
  4. Inspect runtime logs with LOG_LEVEL set to info.

Things to try

Once connected, you can ask your AI assistant things like:

  • List all files in the current project directory
  • Read the contents of README.md in the project root
  • Find files in this project that mention a given function name

Use cases

Accessing local files within a project directory.

Overview

A stdio MCP server configuration example for local filesystem access, using the project directory as its target. The source does not provide specific tool names, client compatibility, or more detailed capability information.

Similar servers

More servers from this repository

Source revision ea10b213686c Data synced 2026-10-19 Read the FMRS scoring method