An MCP server and deterministic DOCX toolkit for AI agents—create, edit, validate, and render reliable Word documents with a local agent playground.
A tiny, meticulous DOCX toolkit for AI agents. Models write; InkPaw keeps the document correct.
InkPaw is a Word document generation, editing, validation, and rendering toolkit for AI agents. It provides an MCP server, a protocol-independent renderer, and a local web Playground for trying the complete agent workflow.
The design principle is simple: the model owns content and structure; deterministic code owns numbering, styles, references, OOXML correctness, and output validation.
The fastest way to experience InkPaw is the local experimental Playground:
Open http://127.0.0.1:8766, enter an OpenAI-compatible base URL, model name, and API key, then ask the agent to create a document.
The Playground shows streaming reasoning and tool activity, renders page previews, supports follow-up edits and templates, and provides DOCX/PDF downloads. It binds to localhost by default and is intended for local evaluation—not direct public deployment. Full page preview requires LibreOffice and poppler-utils. See Playground guide.
InkPaw includes explicit compatibility for DeepSeek thinking-mode Agent Loops: it sends the thinking control, streams reasoningcontent into a separate UI panel, and preserves that reasoning across tool-call continuations as required by the DeepSeek API. The recommended starting model for the Playground is deepseek-v4-flash, which is prefilled with the official https://api.deepseek.com endpoint.
From the project README.
Add the radar badge to your README — it shows your project was picked up by MCP Radar and links to this page:
[](https://mcp.liqiwa.com/s/syouro--InkPaw.html)
MCP server that turns the pi coding agent into a steerable background worker: delegate a task, redirect it mid-run, and keep its context out of your own.
ucsandman/declickCompile an API, an MCP server, a database or a web page into a CLI your agent can run. Ten engines, one output contract, 4.1x less context than raw MCP.
nikships/droid-codex-computer-use🖱️ Let Factory Droid control your Mac with the Codex app's Computer Use tool. One-command install, no approval forms, works with any Codex install path.
leluong141996-dev/Agent-Brain-HubShared, brain-inspired memory for AI agents: 13 brain regions with awake/sleep loops, a live visualizer, agents connected via REST/SDK/MCP, governance & audit, and business-value metrics. Works with C
PerryLink/jevcoreTypeSafe Jev for DeepSeek Harness, the Model Context Protocol, and plain Node: typed judgments instead of prose, offline by default.
sandbaseai/deepseek-harness-handbookIndependent, source-backed handbook for DeepSeek AI's official DeepSeek Harness (dsh): agents, plugins, security, troubleshooting, and runbooks.
The top new MCP servers of the week, every Monday. No spam, unsubscribe anytime.