Look up any MCP server, skill or plugin before your agent uses it.

Each one tells your agent what it does in text you never read, and that text can change after you approved it. Small Print reads new versions as it reaches them, keeps each version it reads, and grades each change by a written rule. No account needed.

Or check everything your agents run

Paste your config in the browseror run this one line in Terminal:

Both send nothing until you say yes. See an example run. Never used a terminal? Step by step. Use VS Code or Cursor? Install Small Print from the extension store, or from Open VSX for Cursor. Want your agent to ask before it uses a server? Add the MCP server.

What you get back

last change · published 2026-11-11

MCP server · npm

@modelcontextprotocol/server-filesystem

Its listing says: MCP server for filesystem access

Tools16 read

Versions25 on record

Changes48 across 9 releases

Worst changemedium

AdvisorieshighGHSA-hc55-p739-j48w highGHSA-q66q-fx2p-7w4m

Versions, newest first · outlined where the record holds a change · each opens the entry page

2.3.0 2.2.0 2.1.0 2.0.1 2.0.0 1.0.0 2026.8.31 2026.7.10 more on the entry page

2.2.0 → 2.3.0mediumpublished 2026-11-11 · 2 descriptions, 12 schemas

mediumdirectory_treetool.schemadrift.tool.schemaThe input form the reader found differs from the one it found in the version before. That is the list of fields the tool accepts and what each field means.

  {    "$schema": "http://json-schema.org/draft-07/schema#",    "properties": {      "depth": {        "description": "How many levels to descend: 1 lists only the entries directly in path. Directories at the last level have no children array. Omit for the whole tree.",        "exclusiveMinimum": 0,        "type": "integer"      },      "excludePatterns": {        "default": [],+       "description": "Patterns for entries to leave out, with their contents, matched against each entry's path relative to path. A pattern with no '*' matches that name at any depth ('node_modules' leaves out every node_modules directory); a pattern with '*' is a glob matched as written, and one ending in '/**' also leaves out the directory it names.",        "items": {          "type": "string"        },        "type": "array"      }    },-   "required": [],    "type": "object"  }

mediumedit_filetool.schemadrift.tool.schemaThe input form the reader found differs from the one it found in the version before. That is the list of fields the tool accepts and what each field means.

  {    "$schema": "http://json-schema.org/draft-07/schema#",    "properties": {      "dryRun": {        "default": false,        "description": "Preview changes using git-style diff format",        "type": "boolean"      },      "edits": {+       "description": "The replacements to make, applied in order, each to the result of the ones before it. If any one has no match, none is written.",        "items": {          "properties": {            "newText": {              "description": "Text to replace with",              "type": "string"            },            "oldText": {              "description": "Text to search for - must match exactly",              "type": "string"            }          },          "required": [            "newText",            "oldText"          ],          "type": "object"        },        "type": "array"      }    },    "required": [      "edits"    ],    "type": "object"  }

mediumlist_directorytool.descriptiondrift.tool.description.changedThe description changed. The added text contains none of the words the instruction, secret or destination rules look for.

Get a detailed listing of all files and directories in a specified path. Results clearly distinguish between files and directories with [FILE] and [DIR] prefixes. This tool is essential for understanding directory structure and finding specific files within a directory.List the files and directories directly inside a path, each marked [FILE] or [DIR]. Use it to see what a directory contains; when you also need each entry's size or want entries sorted by size, use list_directory_with_sizes instead, and for a recursive view use directory_tree. Returns one line per entry. Only works within allowed directories.

and 11 more in this release, on the entry page.

22 more releases compared, on the entry page.

a grade names the rule that matched, not what the author meantthis entry, with its versions and the watch card

The two tags below are examples showing how a grade is written. They are not grades of anything on this page.

advisoryhigh

An advisory is a security report on record. Its own page names its source.

changehigh

A change is a difference in the small print between two versions. It is graded by a written rule. When a model review lowered a high grade or kept it, the entry page says so beside the change.

The record so far

as of 2026-11-14 · cite these numbers, each with its method and date · search the small print

The most recent listing of the sources finished Nov 14, 2:42 AM ET. These figures were counted at Nov 14, 9:30 AM ET, and they include the reads that landed during the day. Every source finished its most recent run on time, though a run can end before it has seen every entry a source lists.

368,226

MCP servers, agent skills and plugins on record, from 20 public sources. 292,862 of them, 79.5%, have been read so far (what counts as one entry)

7,268

security reports on record, each with a page that names its source

1,001,507

separate edits to the small print, from 52,712 entries: one new version can hold many, and each is graded by a written rule, day by day

Four incidents are on record, told from the reporters' own publications. See how each source is read, how we grade, how Small Print differs from scanners and what to know before you roll it out.

From the lab

our own tests, published with the method and the data · every lab

What does your agent pay before it does anything?

$0.25on Sonnet 5.5

$0.50on Opus 5.5

$1.24on Fable 5.1

Every request, without caching, for the tool definitions of one MCP server: hostinger-api-mcp.

$0.14@hasna/todos, every request on Sonnet 5.5; $0.28 on Opus
$0.12@swmansion/argent, every request on Sonnet 5.5; $0.24 on Opus
$0.08@productbrain/mcp, every request on Sonnet 5.5; $0.17 on Opus
ReadWhat the 50 most-downloaded MCP servers cost before your agent does anything
Yoursnpx smallprint cost counts what your own setup sends

Watch what your agents use

Your agents read the small print. Hear when it changes, wherever you already are.

In your inbox: the daily brief comes at the time you choose, with every change graded high or critical, or every change of any grade if you prefer.
In Slack, Discord or on your phone: the same brief can go to a Slack or Discord channel, a webhook, or ntfy for a push notification, on every plan.
In your editor: once you let it check your servers, the VS Code and Cursor extension pops up a notice when one gets a change graded high or critical, or a high or critical security report that covers your version. It checks every six hours, and it needs no account.
On your desktop: with Pro or the 30-day trial, npx smallprint schedule --install checks your instruction files every six hours and shows a notification on a Mac, or on Linux with notify-send, when one changes. On any plan, npx smallprint watch shows one when a server you start through it lists tools that differ from the last time.
For a single package: every entry page has an RSS feed of its changes.

New accounts get 30 days of Pro, no credit card required. After that, the Free plan keeps 25 pins and the daily email. For a team, the pricing page has the prices and the checkout.