LLM_in_Work

skill
Security Audit
Fail
Health Warn
  • License — License: MIT
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Low visibility — Only 5 GitHub stars
Code Fail
  • rm -rf — Recursive force deletion command in LLM_in_Overleaf/install.sh
  • os.homedir — User home directory access in LLM_in_Overleaf/server/cli.js
  • os.homedir — User home directory access in LLM_in_Overleaf/server/config.js
  • process.env — Environment variable access in LLM_in_Overleaf/server/config.js
  • os.homedir — User home directory access in LLM_in_Overleaf/server/models.js
  • process.env — Environment variable access in LLM_in_Overleaf/server/models.js
Permissions Pass
  • Permissions — No dangerous permissions requested

No AI report is available for this listing yet.

SUMMARY

AI writing assistants for Microsoft Word and Overleaf, powered by your local Claude Code or Codex CLI. Review diffs and apply edits in place.

README.md

LLM_in_Work

Bring your local Claude Code and Codex CLI into Word and Overleaf.

English · 简体中文

Select the text you want to improve, describe the change, review the diff, and apply it in the editor you already use. LLM_in_Work brings together two independent writing assistants that use your existing local CLI login.

LLM_in_Word LLM_in_Overleaf
Write in Microsoft Word desktop Overleaf's LaTeX source editor
Review changes Text and table diffs; optional Word tracked changes Per-selection LaTeX diffs; apply as one undoable editor transaction
Edit together Up to 8 paragraphs or tables Multiple non-contiguous selections in one .tex file
Ask questions Use document text as context Use the current .tex file and attached project files
Local connection Office.js add-in → localhost HTTPS → CLI Browser extension → Native Messaging → CLI
Installation Windows and macOS macOS installer for Chrome / Chromium browsers
Interface English and Simplified Chinese Simplified Chinese
Get started Word guide → Overleaf guide →

What you can do

  • Polish, shorten, translate and revise without copying text between an editor and a chat window.
  • Review before applying. Generation produces a preview; you decide when to write it back.
  • Work with multiple targets. Give several selected passages one instruction while keeping the surrounding text intact.
  • Continue the conversation. Refine a draft, ask questions, attach references and revisit local conversation history.
  • Choose your backend. Use Claude Code or Codex CLI, with model selection and supported reasoning levels.

LLM_in_Word

A Word side pane for prose and tables. Preview additions and deletions, apply edits with optional tracked changes, and accept or reject them in Word's Review tab. Formatting is preserved where supported, with an explicit plain-text fallback for unsupported content.

Features, installation and usage · 中文指南

LLM_in_Overleaf

A browser extension for LaTeX writing on overleaf.com and cn.overleaf.com. Collect several selections in one source file, review a replacement for each, then apply them together. Source and filename checks guard against applying stale edits; the whole operation can be undone once in the editor.

Features, installation and usage · 中文指南

Get started

Install Node.js 22.12+ (22.x) or 24+, and install and sign in to at least one local backend: Claude Code or Codex CLI. Then clone this repository:

git clone https://github.com/ZJU-OmniAI/LLM_in_Work.git
cd LLM_in_Work

Choose the Word installation guide or the Overleaf installation guide. Each subproject has its own installer and dependencies; install either or both.

LLM_in_Work/
├── README.md              # Overview and project navigation
├── LLM_in_Word/           # Word add-in, local HTTPS service and installers
└── LLM_in_Overleaf/       # Browser extension and native messaging host

Development and data

Both projects run without additional npm runtime dependencies. Development tests use mock CLIs and synthetic documents. See Contributing for per-project test commands and CI coverage.

The bridge runs locally, but model inference usually connects to the selected provider. Selecting a passage limits the write-back target; it does not mean only that passage is sent as context. Read the data and security notes.

MIT license. An independent ZJU-OmniAI project; not an official Microsoft, Overleaf, Anthropic or OpenAI product.

Reviews (0)

No results found