FilesSAFE
MCP Server for interacting with text-based files (read & write). Written in TypeScript, Node and Hono.dev
Overview
From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.
Stdio MCP server for sandboxed file access — read files, search content, safely edit with checksums, and manage file structure.
Author: overment
Release status (2026-07-27): this repository intentionally pins@modelcontextprotocol/serverand the test-only@modelcontextprotocol/clientto2.0.0-beta.5. These packages implement the2026-07-28release candidate, not a final published protocol. Re-run the final release gate before claiming final conformance.
The server uses the public serveStdio entry with legacy mode set to serve. The SDK chooses the protocol era from the opening exchange and pins one server instance to that stdio connection. Standard output is reserved for MCP frames; diagnostics are written only to standard error.
[!WARNING] This server provides filesystem access to an AI agent. While it's sandboxed to specific directories, always: - Review tool outputs before confirming changes - UsedryRun=trueto preview destructive operations - Keep backups of important files - SetFS_ROOTSto only the directories you want the agent to access
Motivation
Traditional file operations require precise paths and exact content — things LLMs struggle with. This server is designed so AI agents can:
- Explore first — understand directory structure before acting
- Find by name or content — locate files without knowing exact paths
- Edit safely — checksum verification prevents stale overwrites
- Preview changes — dry-run mode shows diffs before applying
- Recover from errors — hints guide the agent to correct mistakes
The result: an agent that can reliably manage your Obsidian vault, documentation, notes, or any text-based file collection.
Features
- ✅ Directory Exploration — tree view with file counts, sizes, timestamps
- ✅ File Reading — line-numbered content with checksums for safe editing
- ✅ File & Content Search — filename search + literal/reg
820921a1602aOBSERVED · 2026-10-07Connect
Built from this server's own package name, version and transport as found in its source — not copied from anyone's documentation, so it cannot drift against a page we do not control.
claude mcp add files-mcp -- npx -y [email protected]
{
"mcpServers": {
"files-mcp": {
"command": "npx",
"args": [
"-y",
"[email protected]"
]
}
}
}Exposed tools (4)
2 read · 1 write · 1 destructive. Blast radius: 1 tool can delete or overwrite — an agent that can be talked into calling a tool can be talked into calling this one.
| Tool | Risk | Description |
|---|---|---|
fs_manage | destructive | Structural operations for files and directories: delete, rename, move, copy, mkdir, stat. |
fs_read | read | Read files or list directories in the sandboxed filesystem. |
fs_search | read | Find files by name and search content within files. |
fs_write | write | Create or update files in the sandboxed filesystem. |
Trust audit
SAFEgrade B · trust 89/100 Nothing in the source contradicts what it says it does. Grade A is reserved for packages that have also passed the behavioural sandbox.
| Layer | What it checks | Result |
|---|---|---|
| L0 | Provenance & inventory | PASS |
| L1 | Static analysis of the code | PASS |
| L2 | Instruction surface (what it tells the agent) | PASS |
| L3 | Class-specific surface | WARN |
| L4 | Behavioural (sandbox) | SKIPPED |
What the source does
- Filesystem
- declared (4 observation(s))
- Network
- none-observed
- Shell
- none-observed
- Dependencies
- not all pinned
- Secrets in source
- none-found
Findings (9)
fs_manage
import { config } from '../../src/config/env.js';import { buildServer } from '../../src/core/mcp.js';import { fsManageTool } from '../../src/tools/fs-manage.tool.js';import { fsReadTool } from '../../src/tools/fs-read.tool.js';const result = await runFsRead({ path: '../../../etc/passwd' });diff, escape-string-regexp, fast-glob, fuzzysort, ignore, picomatch, zod, @biomejs/biome
mcpb/files-mcp.mcpb
Gates applied: no_behavioural_pass, no_license.
820921a1602afull audit observations/trust-audit/mcp-server/iceener__files-1.json · Report an issue / request a re-scanAudit history
Every audit this server has had. A grade with a past is a grade somebody is still checking.
| Date | Source | Verdict | Grade | Score | Change |
|---|---|---|---|---|---|
| 2026-10-07 | 820921a1602a | SAFE | B | 89 | first audit |
Questions
What is the Files MCP server?
MCP Server for interacting with text-based files (read & write). Written in TypeScript, Node and Hono.dev
What tools does Files expose?
4 in total: 2 read-only, 1 that write, and 1 that can delete or overwrite (fs_manage). Every one is listed on this page with its risk.
Is Files safe to connect to an agent?
The audit found nothing in the source that contradicts what it says it does, and graded it B (89/100). Grade A is held back for packages that have also passed a sandboxed behavioural run, which is why a clean server reads B. Separately from the audit: 1 of its tools can destroy data, so scope the token you give it to what you actually need.
What credentials does Files need?
No credential environment variables were found in its source, so it appears to need none.
How current is this page?
The grade is for one exact copy of the source (820921a1602a), read on 2026-10-07. The repository is watched and re-audited when it changes.