GPT-5.6 (Sol / Terra / Luna) is now evaluated on TrustVector โ€” with day-1 independent verification, incl. METR's benchmark-cheating findings.

Read the evaluation
Evaluation record ยท mcp-server-filesystem

MCP Filesystem Server

v2026.7.4

Anthropic

MCPfile-systemlocalmcpmodel-context-protocol
81
Strong
About This MCP

Official MCP reference server providing AI models with controlled access to the local filesystem (file reading, writing, and directory operations). One of seven reference servers still maintained after the 2025-05-29 archival of non-core servers; MCP governance moved to the Agentic AI Foundation (Linux Foundation) on 2025-12-09, latest stable spec 2025-11-25 (2026-07-28 revision at release-candidate stage). Critical for file-based workflows but requires careful security configuration.

Last Evaluated: July 9, 2026
Official Website

Trust Vector Analysis

Dimension Breakdown

๐Ÿš€Performance & Reliability
+
file operation accuracy

File operation testing across multiple platforms

Evidence
MCP Documentation โ€” Direct filesystem operations with high accuracy and reliability
highVerified: 2026-07-09
response time

Latency benchmarking

Evidence
MCP Performance Testing โ€” Local filesystem operations complete in milliseconds
highVerified: 2026-07-09
error handling

Error scenario testing

Evidence
MCP Server Implementation โ€” Comprehensive error handling for permissions, disk space, and invalid paths
highVerified: 2026-07-09
cross platform compatibility

Multi-platform testing

Evidence
MCP GitHub Repository โ€” Supports Windows, macOS, and Linux with platform-specific path handling
highVerified: 2026-07-09
large file handling

Large file performance testing

Evidence
Community Testing โ€” Handles large files efficiently with streaming support
mediumVerified: 2026-07-09
๐Ÿ›ก๏ธSecurity
+
access control

Access control testing

Evidence
MCP Security Documentation โ€” Requires explicit directory allowlist configuration, but enforcement depends on setup
highVerified: 2026-07-09
path traversal protection

Path traversal attack testing

Evidence
MCP Server Code โ€” Built-in path normalization prevents basic traversal attacks
GitHub Advisory Database / Trend Micro disclosure โ€” CVE-2025-53110 (directory containment bypass, CVSS 7.3) and CVE-2025-53109 (symlink bypass, CVSS 8.4) were disclosed by Trend Micro in 2025 and fixed in release 2025.7.1; no newer advisories against this server found as of 2026-07-09 (current release 2026.7.4)
highVerified: 2026-07-09
permission enforcement

Permission boundary testing

Evidence
Security Analysis โ€” Relies on OS-level permissions; AI can request any file within allowed directories
mediumVerified: 2026-07-09
audit logging

Logging capabilities review

Evidence
MCP Logging โ€” Basic operation logging available, detailed audit logs require custom implementation
mediumVerified: 2026-07-09
data exfiltration risk

Data flow analysis

Evidence
Security Review โ€” AI models have read access to all allowed directories; sensitive data exposure risk
highVerified: 2026-07-09
๐Ÿ”’Privacy & Compliance
+
data exposure control

Data flow analysis

Evidence
MCP Architecture โ€” File contents sent to LLM provider; requires careful directory scope configuration
highVerified: 2026-07-09
pii protection

Privacy features assessment

Evidence
MCP Security Guidelines โ€” No built-in PII detection or redaction; responsibility on developer
mediumVerified: 2026-07-09
local processing

Data processing architecture review

Evidence
MCP Architecture โ€” Server runs locally, but file content is transmitted to LLM API
highVerified: 2026-07-09
gdpr compliance

Compliance framework review

Evidence
MCP Documentation โ€” GDPR compliance depends on LLM provider and user configuration
mediumVerified: 2026-07-09
๐Ÿ‘๏ธTrust & Transparency
+
documentation quality

Documentation completeness review

Evidence
MCP Docs โ€” Comprehensive documentation with clear setup instructions and examples
highVerified: 2026-07-09
operation visibility

Logging capabilities assessment

Evidence
MCP Protocol โ€” All filesystem operations visible in MCP message logs
highVerified: 2026-07-09
open source code

Source code review

Evidence
GitHub Repository โ€” Fully open source with MIT license, code auditable
highVerified: 2026-07-09
security guidance

Security documentation review

Evidence
MCP Security Docs โ€” Provides security best practices but could be more comprehensive
mediumVerified: 2026-07-09
โš™๏ธOperational Excellence
+
ease of setup

Setup complexity assessment

Evidence
MCP Quickstart โ€” Simple npm/pip installation with minimal configuration
highVerified: 2026-07-09
resource efficiency

Resource utilization testing

Evidence
Performance Testing โ€” Minimal CPU and memory overhead for file operations
highVerified: 2026-07-09
stability

Issue tracking analysis

Evidence
Community Reports โ€” Stable operation with few reported bugs
highVerified: 2026-07-09
maintenance requirements

Maintenance overhead assessment

Evidence
MCP Server Updates โ€” Regular updates, minimal maintenance required
Anthropic - Donating MCP to the Agentic AI Foundation โ€” filesystem is one of the seven reference servers still actively maintained after the 2025-05-29 archival; MCP governance moved to the Agentic AI Foundation under the Linux Foundation on 2025-12-09
npm - @modelcontextprotocol/server-filesystem โ€” Latest release 2026.7.4 (published 2026-07-04); still listed among the seven maintained reference servers in modelcontextprotocol/servers README (repo ~88,300 stars, not archived)
mediumVerified: 2026-07-09
community support

Community activity analysis

Evidence
GitHub Discussions โ€” Growing community support, relatively new protocol (late 2024)
mediumVerified: 2026-07-09
Strengths
  • +Fast, local file operations with minimal latency
  • +Cross-platform support (Windows, macOS, Linux)
  • +Simple setup and configuration
  • +Open source and fully auditable code
  • +Comprehensive documentation from Anthropic
  • +Efficient handling of large files with streaming support
Limitations
  • !Significant security risk if misconfigured - AI can access all allowed files
  • !No built-in PII or sensitive data detection/redaction
  • !File contents sent to external LLM provider APIs, with potential for accidental data exfiltration
  • !Limited granular permission controls beyond directory allowlists
  • !Audit logging requires custom implementation for compliance needs
  • !SECURITY: CVE-2025-53110 (containment bypass) and CVE-2025-53109 (symlink bypass) fixed in 2025.7.1; versions older than 2025.7.1 remain vulnerable - upgrade to current release
  • !STATUS 2026-07-09: actively maintained reference server (one of 7 retained after the 2025-05-29 archival); latest npm release 2026.7.4 (2026-07-04); governance under the Agentic AI Foundation (Linux Foundation, 2025-12-09)
Metadata
license: MIT
supported platforms
0: Windows
1: macOS
2: Linux
programming languages
0: TypeScript
1: Python
mcp version: 1.0
github repo: https://github.com/modelcontextprotocol/servers
github stars: 88259
first release: 2024-11
maintained by: MCP project (Agentic AI Foundation / Linux Foundation since 2025-12-09)
status: Official - Active
transport types
0: stdio
installation methods
0: npm

Use Case Ratings

code generation

Excellent for code generation workflows requiring file reading and writing

customer support

Limited utility for customer support; mainly for loading reference documents

content creation

Good for content workflows involving file-based templates and outputs

data analysis

Useful for loading data files, but limited analysis capabilities

research assistant

Excellent for accessing research papers, documents, and notes

legal compliance

Useful for document access but requires strict security controls

healthcare

High risk due to PHI exposure; requires careful configuration and HIPAA controls

financial analysis

Risky for sensitive financial data; requires strict directory restrictions

education

Good for accessing educational materials and student work

creative writing

Excellent for managing drafts, outlines, and creative project files