Reimagining AI-assisted writing to let you:
Keep more of your voice and reduce AI slop.
Work alongside the agent in the same live draft, at the same time.
Dynamically reconfigure agency as the writing process changes.
DocWriter is a local editor where you and an AI agent share one live draft. You can work in Markdown, plain text, or LaTeX while the agent reads the same workspace and responds as the draft changes.
You can communicate in the document by writing [[ add a citation to ... ]], selecting text, or replying to a comment. Agent edits appear beside the text for review.
Chat remains available for requests that do not belong at one place in the draft.
DocWriter works with most frontier LLMs and SDKs: Claude Code, Codex, Pi, and others. It uses your existing subscriptions, so if you already have one you do not pay extra.
DocWriter is an open source research project from CMU CSD and UC Berkeley EECS.
Writing is more personal and taste driven than coding, and it is harder to evaluate mechanically. We need interfaces, evaluation methods, and writing practices built for working with AI.
An AI writing harness needs synchronized documents, agent edit tools, skills that encode the writer's voice, and hooks and review passes for repeated work.
A paper or book takes many rounds over months. Writers need ways to redirect agents, keep quality consistent, and review changes without losing their place in the draft.
We want open source methods that help researchers and writers identify repeated AI writing patterns and track how the patterns change across models.
We are looking for user study participants. If you are willing to write and publish a blog post, paper, or other long form document using DocWriter and share your experience, we would love to hear from you.
Open an issue in the DocWriter repository.