siw-charting

An MCP server over a nursing documentation guide: 57 articles built on 121 published court decisions, free, with every decision linked.

Ask it about a situation and it gives you the shape of the note.

draft_note("I got stuck with a needle after a patient")

  template    [DATE] [TIME]: During [PROCEDURE], care was paused due to
              staff sharps injury. Patient remained [CONDITION]...
  do not      "Dirty needle" · "Clean needle" · "Probably fine" · "No risk"
  behind it   Babich v. Waukesha Memorial, 1996
              Barrett v. Danbury Hospital, 1995

Every bracket is a fact you observed. The server writes the shape; it cannot know what happened on your shift and does not pretend to.

It takes no patient information

This is the part to read before anything else.

• Nothing is stored. No log file, no query history, no analytics.
• No network connection is opened. It reads one file from your disk.
• Chart-shaped input is refused, not processed. An MRN, a date of birth, a timed entry, or anything over 200 characters comes back with a refusal instead of an answer.

That is how the server is built. It is not a compliance claim, and none is offered. If you point an assistant at a real patient's chart, what governs that is your employer's policy and whether your model provider has signed a business associate agreement — not a filter written by us. A scrubber does not undo a disclosure that already happened.

Keep the patient out of the question and the question stays easy.

Install

No packages, no build step. Python 3.9 or newer.

curl -O https://shiftiswild.com/notes/siw_charting.py
curl -O https://shiftiswild.com/notes/corpus.json
python3 siw_charting.py --check

Both files live in the same folder. --check runs the whole self-test, including the refusals, and needs no client.

Point a client at it

Claude Code

claude mcp add siw-charting -- python3 /full/path/to/siw_charting.py

Claude Desktop — ~/Library/Application Support/Claude/claude_desktop_config.json on macOS, %APPDATA%\\Claude\\claude_desktop_config.json on Windows.

Cursor — ~/.cursor/mcp.json, or .cursor/mcp.json per project.

Config paths move between versions; the block inside is the same everywhere:

{
  "mcpServers": {
    "siw-charting": {
      "command": "python3",
      "args": ["/full/path/to/siw_charting.py"]
    }
  }
}

Tools

tool	what it gives you
draft_note	template for the note, wording to avoid, decisions behind it
next_step	what to do besides writing it down, in order
search_charting	find articles by situation
read_article	a whole article by slug
find_phrases	stronger wording for a situation
words_to_avoid	wording that reads badly later, and what to write instead
charting_cases	the decisions themselves — court, year, link
list_sections	everything, grouped, in reading order

It also ships prompts: clients that support them show two ready conversations — Help me write this note and What do I do now — and both carry the rules with them: never ask for patient identifiers, never invent a finding, name the decision so the nurse can check you.

Where the content comes from

Every rule in the guide stands on a published decision on CourtListener, linked so you can read it yourself. Where only a docket entry was available and not the opinion, the case is left out — a decision nobody can open is a citation, not evidence.

Written by a medical-surgical registered nurse, ten years, day shift. Not legal advice and not clinical advice.

• The guide: https://shiftiswild.com/notes/
• Every decision in one page: https://shiftiswild.com/notes/cases/
• Machine-readable manifest: https://shiftiswild.com/notes/mcp.json
• The corpus as JSON: https://shiftiswild.com/notes/corpus.json

Licence

Code — MIT. Content in corpus.json — CC BY 4.0: use it, quote it, build on it, with a link back to https://shiftiswild.com/notes/.

The source comments are in Russian; the project is written that way. The tool descriptions, prompts and everything a user sees are in English.