paper-format-agent
A local DOCX formatter for theses and papers that also provides an optional MCP server with tools for formatting, extracting rules, and scoring papers.
README
<!-- mcp-name: io.github.zxyasfas/paper-format-agent -->
Paper Format Agent
中文说明 | English
A local DOCX formatter for theses and papers.
It changes fonts, spacing, indents, headings and captions to match a format guide. It does not rewrite the paper. Before saving, it compares a fingerprint of the body and table text from before and after the run. If the text changed, it aborts instead of writing the formatted file.
Everything runs on your machine. Nothing is uploaded.
the content check
Report fields from a real run (--engine python):
{
"content_fingerprint_before": "793e6533fd670418141d11fdcf014be19750408129ecff8b1b78a2641a3786db",
"content_fingerprint_after": "793e6533fd670418141d11fdcf014be19750408129ecff8b1b78a2641a3786db",
"content_changed": false,
"content_guard_enforced": true
}
The two hashes should match. If they don't, the formatter exits with
content guard failed and the formatted DOCX is not written.
What the check covers: body paragraphs and tables, with whitespace and stray
bullet characters normalized before comparing. Headers and footers are out of
scope because the formatter sets those on purpose. Use --engine python when
the fingerprint must cover the final saved DOCX; the other engines run a local
post-processor after the check, e.g. to refresh the table of contents.
To watch the guard trip, run python tools/demo_content_guard.py. It formats
a synthetic paper, then repeats the run with the styling step patched to edit
one sentence of the in-memory document. The second run aborts without writing
the DOCX.
docs/BENCHMARK.md tracks which authored strings survive a run in small synthetic fixtures, and lists the known gaps.
install and run
pip install -r requirements.txt
python -m paper_format_agent.cli \
--format-file "format_guide.docx" \
--paper-file "paper.docx" \
--out-dir "./output" \
--engine python
There is also a GUI (python run_gui.py), a batch mode, and JSON template
packs for Chinese thesis, journal and IEEE-style formats. See
docs/USAGE.md.
limits
Early project. Ordinary paragraphs, headings and tables work better than
equations, footnotes and complex layouts. With --strict-required-sections,
checks may include char_below_min or blank_page_risk. Do not use it as the
only check before a real submission. Keep your original file.
Reference-list formatting is out of scope here: that part is mechanical rules, so it lives in a separate tool, gbt7714_check, which lints reference entries against GB/T 7714.
why I made this
I did not want to upload an unfinished thesis to a formatting website, and I wanted a way to check that formatting did not touch the text. So the default engine runs locally and refuses to save if the text changed.
agents
The repo doubles as an installable agent skill (SKILL.md), and the same pipeline is available as an optional MCP server (Python 3.10+):
pip install "paper-format-agent[mcp]"
paper-format-agent-mcp
Tools: format_paper, extract_format_rules, score_paper. Client config in
docs/MCP.md.
contributing
Small PRs are welcome. Adding a synthetic test for a school or journal formatting rule is a good place to start: docs/CONTRIBUTOR_TASKS.md, CONTRIBUTING.md. Notes on the pipeline are in docs/ARCHITECTURE.md.
Do not commit real papers, private school templates, reviewer comments, API keys or generated output. Synthetic fixtures only.
license
MIT. See LICENSE.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
Neon Database
MCP server for interacting with Neon Management API and databases
E2B
Using MCP to run code via e2b.
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.