precis-mcp 3.0.0

Unified document MCP — read, write, search, annotate any structured document

Precis

Unified document MCP — read, write, search, and annotate any structured document.

Four tools

search(query, top_k?, scope?)
get(id, grep?, depth?)
put(id, text?, mode?, tracked?, note?, link?)
move(id, after)

URI grammar

scheme:path[~selector][/view[/subview]]

Schemes

• file: — on-disk files, extension determines handler (.docx, .tex, .md, .txt)
• paper: — acatome paper store (pre-ingested PDFs)
• todo: — task management with state machine

Examples

read('paper:')                              # list all papers
read('paper:miller2023foo')                 # overview + abstract
read('paper:miller2023foo/toc')             # structure
read('paper:miller2023foo~38')              # chunk 38
read('paper:miller2023foo/cite/bib')        # BibTeX citation
read('file:planning.docx')                  # table of contents
read('file:planning.docx~KR8M2')            # node by slug
put('file:planning.docx~KR8M2', text='Revised text.', mode='replace')

Output markers

=  verbatim (safe to quote)
~  derived (keywords/summary — not quotable)
%  annotation (user note/comment)

Installation

pip install precis-mcp            # core + markdown + plaintext + latex
pip install precis-mcp[word]       # + DOCX support
pip install precis-mcp[paper]      # + paper store support
pip install precis-mcp[all]        # everything

Plugin system

Register new schemes or file types via entry points:

[project.entry-points."precis.schemes"]
chem = "my_plugin:ChemHandler"

[project.entry-points."precis.file_types"]
".sdf" = "my_plugin:SDFParser"