SoloMD: A Local-First Markdown Editor That Lets Claude and Cursor Drive Your Vault
A markdown editor, and the bridge to your LLM. Local-first, MIT, ~15 MB. Bundled MCP server lets Claude Code / Codex / Cursor drive your vault directly. 14 AI providers BYOK.
At a glance
- What is it?
- SoloMD is a Tauri-based Markdown editor with a bundled MCP server, letting AI clients like Claude Code and Cursor read and write your notes directly. It is local-first, MIT-licensed, and weighs about 15 MB, but its agent-centric design trades some simplicity for a steeper setup.
- Who is it for?
- Adopt SoloMD if you want a Markdown editor that doubles as a local MCP endpoint for AI agents, especially if you already use Claude Code or Cursor and want your notes to stay on your machine. Skip it if you need a simple, distraction-free editor without agent integration, or if you rely on Windows 7/8.1.
- Can I use it commercially?
- Yes. MIT is a permissive licence: you can use, modify and sell software built on it, as long as you keep its copyright and licence notices.
- Is it still maintained?
- Yes. The repository last received commits 1 day ago.
- What is it written in?
- Mainly TypeScript, according to GitHub's language statistics.
Answers come from the project's GitHub data, last synced on September 14, 2026, and from our analysis. They are not legal advice.
DEEP OPEN-SOURCE ANALYSIS
The Problem: Editors and Agents Live in Separate Worlds
Most Markdown editors are passive tools. You type, you save, you close. AI agents like Claude Code or Cursor, on the other hand, work best when they can read and write files directly. The gap between these two worlds is real: you either copy notes into a chat window or give an agent access to a whole directory with no structure. SoloMD tries to close that gap by making the editor itself the bridge. It is built for people who keep a folder of Markdown notes and want an AI agent to act on that folder without exporting or duplicating content. The README is explicit about the audience: the editor has a 'first-class agent surface inside' and an MCP endpoint that Claude Code or Cursor can drive from outside. If you are a developer, researcher, or knowledge worker who already uses AI coding tools, this is aimed at you.
How It Works: MCP Server, Local-First, and the Same .md Files
The core mechanism is the bundled MCP server, a separate crate at `mcp-server/`. It runs against any folder of Markdown files, and the README claims you do not need SoloMD itself installed to use the server. The MCP server exposes your vault to AI clients through the Model Context Protocol, which is how Claude Code, Cursor, and similar tools talk to external data sources. The data flow is simple: the AI client sends a request to the MCP server, the server reads or writes the Markdown files in the specified workspace, and the results go back. The server is path-traversal guarded and listens on no network port, so it only sees what you point the workspace at. This is a local-first design: notes, AI keys, embeddings index, and git history all stay on your machine. There are no SoloMD-hosted servers, and the license is MIT. The editor itself is built on Tauri 2, Vue 3, and CodeMirror 6, which explains the small footprint: the macOS dmg is about 32 MB, and the core is around 15 MB.
Getting Running: Commands, Config, and Multi-Vault Setup
Installation is straightforward. On macOS, you can use Homebrew: `brew install --cask zhitongblog/solomd/solomd`, or download the universal dmg. Windows users get an MSI or a portable zip, and there is a PowerShell one-liner: `irm https://solomd.app/install.ps1 | iex`. Linux has AppImage, deb, and rpm packages, plus an AUR package. Once installed, the key command for agent integration is `solomd mcp-config`, which prints a JSON snippet for your AI client. The snippet points to the bundled `solomd-mcp` binary and includes a `--workspace` argument. For multiple vaults, you repeat the flag: `--workspace /Users/me/Documents/SoloMD --workspace /Users/me/Documents/work-notes`. There is also a CLI command for direct prompts: `solomd agent "rewrite this week of dailies into a weekly review and commit it"`. For developers, the repository builds with `pnpm install` and `pnpm tauri dev`, and the MCP server is a separate crate. Linux needs `libdbus-1-dev` for the keychain backend.
The Knowledge Graph Layer: More Than a Text Editor
Version 4.6 added a knowledge-graph layer that the README says 'the structural half Obsidian and Tolaria had and SoloMD didn't'. This includes a properties inspector for editing YAML frontmatter with type, date, status, and relation pickers. The writes are line-surgical, meaning they preserve comments, inline arrays, and quoting byte-for-byte. There is a type-driven sidebar where a note with `type: Project` becomes a collapsible section with custom icons and columns. Typed relationships like `belongs_to` and `related_to` get automatic inverses, resolved server-side for large vaults. There is also a per-note relationship graph called 'Neighborhood', saved filtered views stored as `.solomd/views/*.yml`, and an inbox workflow with `⌘E`. This is a significant functional addition, but it also means SoloMD is not just a simple editor anymore. The knowledge graph adds complexity to the file format, even if it stays in Markdown. The README mentions tldraw whiteboards that are Markdown-backed and cross-compatible with Tolaria's on-disk format, but the lazy-loading hint suggests performance is a consideration.
Limitations and Failure Modes: When SoloMD Is the Wrong Tool
The most obvious limitation is the Windows 7/8.1 exclusion. The README is blunt: 'Windows 7/8/8.1 cannot be supported' because the Rust toolchain requires Windows 10 and Microsoft froze WebView2 at version 109 on Windows 7 with no security updates. There is no legacy build. That is a hard boundary for anyone stuck on older Windows. Another limitation is the MCP server's scope. It only sees what you point the workspace at, which is a security feature, but it also means you have to manage multiple `--workspace` flags if your notes are spread across folders. The path-traversal guard is mentioned, but the README does not detail how it handles symlinks or nested paths. If you have a vault with deeply nested links or non-standard Markdown extensions, you may need to test the server's behavior yourself. Also, the agent integration is the selling point, but it is also a failure mode: if you do not use Claude Code, Cursor, or another MCP client, the bundled server is useless. You are left with a Markdown editor that competes with Typora and Obsidian, but without their mature plugin ecosystems.
Alternatives: Obsidian, Typora, and Tolaria Compared
The README itself points to a comparison page against Obsidian, Typora, and Tolaria. Obsidian is the obvious alternative: it has a mature knowledge graph, plugins, and a large user base, but it is not open source and its AI integrations are less direct. Typora is a WYSIWYG editor that is simpler and more focused, but it has no built-in agent surface or MCP server. Tolaria appears to be a newer project with an on-disk format for whiteboards that SoloMD claims to be cross-compatible with. The key difference is the MCP server. SoloMD is the only one of these that ships a bundled MCP endpoint as a first-class feature. Obsidian has community plugins for AI, but they are not native. Typora has no agent integration at all. Tolaria may have similar ideas, but the README suggests SoloMD is catching up to its knowledge graph. If you need an agent to drive your notes directly, SoloMD has a structural advantage. If you just want to write Markdown, any of the three will do.
Maintenance, License, and Upgrade Cost
The project is MIT-licensed, which means you can fork and modify it freely, but you get no warranty. The README shows a single maintainer, xiangdong li, with contact via Telegram and WeChat. The release history shows frequent updates: v4.11.20, v4.11.19, and v4.11.17 within a week in August 2026. That suggests active maintenance, but it also means upgrade cost is real. Each release may change the MCP config format or the knowledge graph schema, and you will need to track those changes if you rely on the agent integration. The repository layout includes a `docs/roadmap.md` and a build log, which helps, but you should budget time for reading release notes. The MCP server is a separate crate, so upgrading the editor does not necessarily upgrade the server, and vice versa. You also need to manage your AI provider keys yourself; the README mentions 14 AI providers with BYOK, but there is no hosted key management. If you use the multi-workspace federation, you will need to update the JSON config every time you add or remove a vault.
Editorial conclusion
Adopt SoloMD if you want a Markdown editor that doubles as a local MCP endpoint for AI agents, especially if you already use Claude Code or Cursor and want your notes to stay on your machine. Skip it if you need a simple, distraction-free editor without agent integration, or if you rely on Windows 7/8.1. Before adopting, verify the MCP server's path-traversal guards in your own vault structure, test the multi-workspace federation with your actual folder layout, and confirm that the bundled `solomd-mcp` binary works with your AI client's version. The project is actively maintained with regular releases, but its agent-native approach is a deliberate trade-off: you get a powerful bridge, but you must trust the MCP endpoint and manage multiple configuration files.
Community notes