pyVBAReference

A complete, machine-generated reference of the VBA (Visual Basic for Applications) object models for Excel, PowerPoint, Word and Access, plus the nine shared COM type libraries every host can reference. Every public type is exported as both human-readable Markdown and machine-readable JSON.

Layout

All generated data lives under reference/ to keep the repo root clean:

reference/
  agentic_llm_primer.md  guide for AI agents (schema + grounding)
  index.json             master catalog of every library and type
  members.json           member name -> the types that define it
  <library>/
    md/     one .md per type   + _index.md
    json/   one .json per type + _index.json
  consolidated/          the same data folded into one set of files per app
    <app>.json / .md             every type
    <app>_constants.json / .md   every enumeration and module constant
    <app>_properties.json / .md  every property of every object

Both md/ and json/ hold the same data: signatures, return types, parameter lists, property access modes, enum values, remarks, and examples - introspected from the registered COM type libraries and enriched with descriptions from Microsoft Learn.

Libraries

Host applications:

Shared libraries, referenceable from any host:

VBA language built-ins (MsgBox, Format, CStr, vbCrLf, ...) live in the vba library, grouped into modules such as Interaction and Strings.

A type name that exists in several hosts (Application, Range, Font, ...) resolves to the Excel one first; pass a library to pick another (vba.get_type("Application", "word")).

Finding something

• Don't know which type owns a member? Check reference/members.json - it maps every member name to the types that define it.
• Want the full catalog? See reference/index.json - all libraries and types.
• Know the type name? Open reference/<library>/md/<TypeName>.md.
• Browsing one library? Start at reference/<library>/md/_index.md.
• A global function (e.g. MsgBox)? It's in a module under reference/vba/ - MsgBox is in reference/vba/md/Interaction.md.
• A constant's value (e.g. xlCSV)? See the enum file, e.g. reference/excel/md/XlFileFormat.md.
• Want one file instead of a folder tree? See reference/consolidated/ below.

Consolidated exports

reference/consolidated/ holds the same data folded into one set of files per application - Excel, PowerPoint, Word, Access, and shared for the libraries all four have in common. Each set is three pairs of files:

The two CSVs are flat tables for importing into a spreadsheet. Properties, as Object,Property,Property split,Type,Access:

Application,ActiveCell,Active Cell,Range,R_O
Worksheet,UsedRange,Used Range,Range,R_O
Worksheet,Name,Name,String,V

Property split is the same name split at capitals, for reading rather than calling. Access is R_O read-only, V settable, W_O write-only.

Constants, as Owner,Kind,Constant,Value,Description:

XlFileFormat,Enumeration,xlCSV,6,CSV
XlFileFormat,Enumeration,xlWorkbookDefault,51,Workbook default
Constants,Module,vbCrLf,Chr(13) & Chr(10),

Value is written as a VB6 literal, so it drops straight into generated code: bare numbers, quoted strings, and Chr() for the handful of constants whose value is a control character. 26,543 property rows and 12,995 constant rows across the ten files.

So the whole Word object model is one 3.6 MB JSON file or one 1.8 MB Markdown file; every Excel constant is a 175 KB Markdown file. Nothing here is new data - it is the per-type files concatenated, for grep, download, or feeding to a model. Start at reference/consolidated/_index.md.

Regenerating

Requires Windows with the relevant Office applications installed and pywin32 (pip install -e ".[generate]"):

.venv\Scripts\python.exe scrape_excel_object_model.py

Flags: --no-enrich (signatures only, skip Microsoft Learn text), --refresh-docs (force re-download of the documentation corpus), and --only word,powerpoint (rebuild just those libraries; the master indexes and consolidated exports still cover everything else).

The consolidated exports are rebuilt at the end of every run. To rebuild only them - no Office or pywin32 needed, just the generated JSON:

python consolidate_reference.py

For AI agents

See agentic_llm_primer.md for the JSON schema and guidance on grounding VBA code against this reference.

Python library

The same data is exposed as an installable, typed Python package, vba_reference. The JSON is bundled into the wheel, so an installed copy is self-contained; in this repo it reads the generated folders directly.

pip install -e .          # from this repo (editable)
# or: pip install vba-reference

import vba_reference as vba

vba.library_names()                          # ['excel', 'powerpoint', 'word', ...]
ws = vba.get_type("Worksheet")               # TypeDoc (case-insensitive)
print(ws.remarks)
protect = ws.member("Protect")               # Member
[(p.name, p.optional) for p in protect.parameters]

vba.get_type("Document", "word")              # disambiguate by library
vba.locate_type("Application")                # every host that defines it
vba.find_members("MsgBox")                    # -> [MemberRef(library='vba', type='Interaction', ...)]
vba.find_members("SaveAs")                    # every type that defines SaveAs
vba.get_constant("XlFileFormat", "xlCSV").value   # 6
vba.get_constant("WdSaveFormat", "wdFormatPDF").value  # 17

Command-line interface (vba-ref once installed, or python -m vba_reference):

vba-ref libs                     # list libraries and type counts
vba-ref where MsgBox             # where a member/type is defined
vba-ref type Worksheet           # full type entry
vba-ref member Worksheet Protect # one member with parameter docs

Security

Every push to main, every pull request and every release is scanned with CodeQL, Semgrep, ClamAV and YARA-X, and again daily. The build fails on anything that is not a reviewed, documented exception. Each release carries its security report as an attached file. To report a vulnerability privately, and for what is scanned and how, see SECURITY.md.

License and attribution

The code in this repository is MIT licensed - see LICENSE. The attribution for the documentation text in the data is in THIRD_PARTY_NOTICES.md, and both files ship in the package.

The generated data is two things with two origins. Type names, member signatures, parameter lists, property access modes and enumeration values are introspected from the COM type libraries themselves. The prose - descriptions, remarks and code examples - comes from MicrosoftDocs/VBA-Docs, published by Microsoft Corporation under CC BY 4.0 with its code samples under the MIT License. It is redistributed here under CC BY 4.0 with attribution to Microsoft.

This project is not affiliated with or endorsed by Microsoft.