About docsdiff
docsdiff is a local tool for reviewers and teams to detect whether code and documentation changed together. It classifies changed paths and reports drift by area without reading repository files.
Key features
- 01Classifies files by path and extension
- 02Reports drift status per area
- 03Lists drifted areas and file counts
- 04Validates change-set input
Use cases
- 01Reviewing whether a change needs documentation updates
- 02Checking documentation-only changes for missing code updates
MCP support
The MCP server exposes the `check_docs_drift` tool, which accepts changed file paths and optional area labels and returns per-area statuses, drifted areas, and counts.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
Tools
1Saved tool list · updated 1w ago
| Parameter | Type | Required | Description |
|---|---|---|---|
| changes | array | Yes | Changed files to classify (at least one) |
Details
From the project READMEdocsdiff
Code and its documentation drift apart quietly. A pull request rewrites a module and leaves the guide that describes it untouched; a docs page is corrected for behaviour the code never had. Reviewers catch some of it, but the signal is buried in the file list. docsdiff makes that signal explicit: you hand it the changed files of a change set, and it reports, per area, whether code changed without docs, docs changed without code, the two moved together, or nothing relevant changed at all.
Every file is classified from its path alone — no contents are read, no repository is opened. The checks run in a fixed order and the first match wins:
- docs — the path starts with
docs/or contains a/docs/segment, the basename (lowercased) starts withreadme, or the extension ismd,mdx, orrst. - code — the extension is one of
ts,tsx,js,jsx,mjs,cjs,py,go,rs,java,rb,php,c,h,cpp,hpp, orcs. - other — everything else, including files with no
extension and hidden files such as
.gitignore.
The extension is the text after the last . in the basename,
lowercased, so PAGE.MD is docs and App.TS is code just as
their lowercase spellings are. Because the docs check runs
first, a README inside a code directory is docs even when its
name carries a code extension: src/readme.py is docs, not
code.
Each file may carry an area — a module, package, or any
other grouping the caller uses. Files without an area are
reported under the literal area name "(global)". For every
area that appears, the report counts its code and docs
changes and assigns one status:
code_without_docs— code changed, docs did notdocs_without_code— docs changed, code did notin_sync— both changedno_relevant_changes— neither changed (only other files touched the area)
The result also lists the drifted areas — those in either
drift status — sorted ascending, and overall counts of files,
code, docs, and other. Areas are sorted ascending in code-unit
order, with "(global)" sorting by its literal characters.
Malformed input — a missing, empty, or non-array changes
list, an entry that is not an object, a missing, empty, or
duplicated path, an area that is present but empty or not
a string — is a validation error naming the offending path
(for example changes[1].area must be a non-empty string),
and no partial report is produced. The computation is pure
and local: no network, no configuration, no files read beyond
the input you pass in.
CLI
npx @reneturcios/docsdiff check --input args.json
npx @reneturcios/docsdiff --version
args.json contains the full arguments object:
{
"changes": [
{ "path": "src/api/handler.ts", "area": "api" },
{ "path": "src/api/auth.ts", "area": "api" },
{ "path": "docs/guide.md" },
{ "path": "assets/logo.png", "area": "api" }
]
}
The report is printed as JSON:
{
"areas": [
{ "area": "(global)", "code_changed": 0, "docs_changed": 1, "status": "docs_without_code" },
{ "area": "api", "code_changed": 2, "docs_changed": 0, "status": "code_without_docs" }
],
"drifted": ["(global)", "api"],
"counts": { "files": 4, "code": 2, "docs": 1, "other": 1 }
}
Exit codes: 0 on success, 1 on a runtime error (including
an unreadable or invalid input file, or invalid input), 2
on a usage error.
MCP server
npx -y @reneturcios/docsdiff mcp
MCP client config:
{
"mcpServers": {
"docsdiff": {
"command": "npx",
"args": ["-y", "@reneturcios/docsdiff", "mcp"]
}
}
}
Tools
-
check_docs_drift— Check a change set for documentation drift per area.changes(array, required, at least one) — changed files as{path, area?}objects;pathis unique across the set, andareagroups the file under a named area instead of"(global)"
Returns
{areas, drifted, counts}.
Docker
docker run -i ghcr.io/cloudcorprecords/docsdiff
docker run --rm -v "$PWD:/w" ghcr.io/cloudcorprecords/docsdiff check --input /w/args.json
Development
npm install
npm run build # tsc -> dist/
npm test # tsx --test test/*.test.ts (pure local, no network)
npm run dev # run the CLI from source
License
MIT — Copyright (c) 2026 Rene Turcios
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.