This is a smart move for agentic workflows. The biggest pain with using LLMs on docs is the clunky API access to tools like Notion. Treating the knowledge base as a structured filesystem via MCP makes it feel like a legitimate dev tool rather than just another wiki.
Man, where was this a couple months ago? I was managing a janky review workflow in my notion spaces to prevent Claude from bashing my knowledge base in those times it starts to circle; so much nicer to just know there's a state to roll back to when those states arise.
The "revert to this commit" + commit based diff viewer hopefully helps the next time this comes up for you. If you have questions don't hesitate to reach out.
Moxn is a collaborative editor with git-like version control for documents and a filesystem interface for agents.
Even before AI, I wanted docs to behave more like repos for two reasons:
- With multiple users live-editing in the same doc, the content I was working on would bounce around in the UI.
- I'm naturally iterative and I don’t one-shot prose well.
Collaborating with AI added two more pain points:
- Multiple agents live-editing the same doc leads to clobbers and deadlocks or file replication and name-based versioning.
- Lack of granular agent-specific permissions.
The common options helped, but came with tradeoffs:
- Git + markdown works well for agents and individual work on a machine, but teams lose multiplayer editing.
- Notion and Google Docs work better for teams, but agents have to go through APIs that weren't designed for filesystem-style exploration. The versioning of the docs is revision-based, not branch-based, and search is flat.
Moxn is an attempt to combine the git + markdown and Notion/Google Docs worlds to support iterative development and collaboration on content.
For humans: a multimodal editor with comments, backlinks and real-time multiplayer. Humans can set branch and section-specific AI permissions.
For agents: a filesystem interface with `find`, `search` and `read` tools for content organized by paths, tags, and databases.
The part I care the most about is the version control. Docs have branches, commits and merge requests. Merges are common-ancestor aware and conflicts resolve at the section grain. Edits flow into the CRDT (Yjs/TipTap), get flushed to mutable draft records, and frozen on commit. Edits have `git staged` semantics and agents default to auto-commit.
I use different workflows for different types of work with Moxn.
- Coding: I push complex plans with Mermaid and bug reproductions with screenshots via Playwright to review and leave comments. I have a post-commit hook that runs Claude Code to auto-update docs on a branch with the same name as my code edits and create a changeset, which is a coordinated collection of merge requests.
- Content: I fan agents out on branches to propose candidates (still an MLE at heart), read the diffs, cherry-pick chunks that I like, drop into the editor to write and loop. The launch post https://moxn.dev/blog/launch was written that way.
Rough edges I want to call out:
- Editor cold-start and polish aren't at Notion/Obsidian levels yet.
- Full-text search only covers the default branch today, and doesn't hit metadata or multimodal content yet.
- No audit logging or other enterprise features yet.
There’s a free plan with no credit card required. The blog post has screenshots, and the docs have a video walkthrough including a Playwright bug reproduction doc.
I'd especially love feedback from anyone using agents against internal docs or knowledge bases who has hit retrieval, coordination or collaboration problems.
This is a smart move for agentic workflows. The biggest pain with using LLMs on docs is the clunky API access to tools like Notion. Treating the knowledge base as a structured filesystem via MCP makes it feel like a legitimate dev tool rather than just another wiki.
Man, where was this a couple months ago? I was managing a janky review workflow in my notion spaces to prevent Claude from bashing my knowledge base in those times it starts to circle; so much nicer to just know there's a state to roll back to when those states arise.
The "revert to this commit" + commit based diff viewer hopefully helps the next time this comes up for you. If you have questions don't hesitate to reach out.
This is great; it's honestly so f'in annoying working with documents with multiple users AND LLMs
Thank you!
Moxn is a collaborative editor with git-like version control for documents and a filesystem interface for agents.
Even before AI, I wanted docs to behave more like repos for two reasons:
- With multiple users live-editing in the same doc, the content I was working on would bounce around in the UI.
- I'm naturally iterative and I don’t one-shot prose well.
Collaborating with AI added two more pain points:
- Multiple agents live-editing the same doc leads to clobbers and deadlocks or file replication and name-based versioning.
- Lack of granular agent-specific permissions.
The common options helped, but came with tradeoffs:
- Git + markdown works well for agents and individual work on a machine, but teams lose multiplayer editing.
- Notion and Google Docs work better for teams, but agents have to go through APIs that weren't designed for filesystem-style exploration. The versioning of the docs is revision-based, not branch-based, and search is flat.
Moxn is an attempt to combine the git + markdown and Notion/Google Docs worlds to support iterative development and collaboration on content.
For humans: a multimodal editor with comments, backlinks and real-time multiplayer. Humans can set branch and section-specific AI permissions.
For agents: a filesystem interface with `find`, `search` and `read` tools for content organized by paths, tags, and databases.
The part I care the most about is the version control. Docs have branches, commits and merge requests. Merges are common-ancestor aware and conflicts resolve at the section grain. Edits flow into the CRDT (Yjs/TipTap), get flushed to mutable draft records, and frozen on commit. Edits have `git staged` semantics and agents default to auto-commit.
I use different workflows for different types of work with Moxn.
- Coding: I push complex plans with Mermaid and bug reproductions with screenshots via Playwright to review and leave comments. I have a post-commit hook that runs Claude Code to auto-update docs on a branch with the same name as my code edits and create a changeset, which is a coordinated collection of merge requests.
- Content: I fan agents out on branches to propose candidates (still an MLE at heart), read the diffs, cherry-pick chunks that I like, drop into the editor to write and loop. The launch post https://moxn.dev/blog/launch was written that way.
Rough edges I want to call out:
- Editor cold-start and polish aren't at Notion/Obsidian levels yet.
- Full-text search only covers the default branch today, and doesn't hit metadata or multimodal content yet.
- No audit logging or other enterprise features yet.
There’s a free plan with no credit card required. The blog post has screenshots, and the docs have a video walkthrough including a Playwright bug reproduction doc.
I'd especially love feedback from anyone using agents against internal docs or knowledge bases who has hit retrieval, coordination or collaboration problems.