SkillAgentSearch skills...

solomd

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.

Install / Use

claude mcp add zhitongblog -- npx -y github:zhitongblog/solomd

If the server publishes to npm under a different name, use that package instead — check the repo README.

About this skill
🔌

MCP Server

Model Context Protocol server

Quality Score

93/100

Supported Platforms

Claude Code
Claude Desktop
Cursor
OpenAI Codex

Tags

Our assessment of solomd

solomd scores 93/100 on our quality scale, 206th of 932 AI & Machine Learning skills we index (top 23%).

Its MCP Server is 20 KB long, well organised into 22 sections with 11 code examples: a thorough specification that gives an agent plenty to work with.

With 1,422 GitHub stars, it is one of the more widely adopted skills in the catalogue.

Substance
30/30
Structure
20/20
Description
15/15
Adoption
13/20
Freshness
15/15

Maintenance, license and trust

  • The repository was last updated today, so solomd is actively maintained.
  • Our last check on 2026-09-06 found the source still online.
  • It is released under the MIT license, a permissive license that allows use, modification and commercial use with attribution.
  • Its trust signals score 100/100, with no cautions. These come from repository metadata, not a code audit — read the skill file before letting an agent act on it.

Safety scan

No issues found

Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands (3 minor notes below). An AI review of the same text found nothing harmful.

  • noteInstalls by piping a downloaded script into a shellline 139
    curl -fsSL https://raw.githubusercontent.com/zhitongblog/solomd/main/scripts/install-cli.sh | bash
  • noteInstalls by piping a downloaded script into a shellline 163
    curl -fsSL https://solomd.app/install.sh | bash
  • noteInstalls by piping a downloaded script into a shellline 171
    irm https://solomd.app/install.ps1 | iex

AI review by kimi-k2.7-code on 2026-10-13. Automated pattern scan on 2026-10-13. It catches known dangerous patterns, not every risk — read a skill before letting an agent act on it.

solomd compared with similar skills

All 4 of these similar skills score higher than solomd; compare them before choosing.

SkillScoreStarsUpdatedFormat
solomd (this skill)by zhitongblog931.4ktodayMCP Server
Agent-Reachby Panniantong100105.7k4d agoCLAUDE.md
claude-memby thedotmack100104.1ktodayCLAUDE.md
Understand-Anythingby Egonex-AI10088.8k3d agoCLAUDE.md
headroomby headroomlabs-ai10078.2ktodayCLAUDE.md

Frequently asked questions

How do I install solomd?
Run claude mcp add zhitongblog -- npx -y github:zhitongblog/solomd. The install tabs above show the steps for each supported agent.
Which AI agents does solomd work with?
It is written for Claude Code, Claude Desktop, Cursor and OpenAI Codex, as a MCP Server file. Other agents that read the same format can often use it too.
Is solomd safe to use?
Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands (3 minor notes below). An AI review of the same text found nothing harmful. It is MIT-licensed and scores 100/100 on trust signals. Skills are instructions an agent will follow, so read the file before installing it and do not approve commands you do not understand.
Is solomd still maintained?
The repository was last updated today, so solomd is actively maintained.

SoloMD

The editor where agents live.

Latest release License: MIT Downloads GitHub stars Website

🌐 中文 · 日本語 · 한국어 · Deutsch · Français · Español · Português · Italiano · Polski · Nederlands · Türkçe · Svenska · Українська · 🪞 Gitee mirror → (faster downloads from China)

Download the latest release · Launch post · How we built it · Website · Security

⭐ If SoloMD earns a place on your machine, star it on GitHub — it's free, made by one developer, and stars are how other people find it.

SoloMD editor

🤖 Built and maintained by Claude

SoloMD is written, tested and released almost entirely by Claude Code, Anthropic's coding agent, working under a human maintainer who sets the direction and decides what ships. Issues and pull requests are triaged and answered by Claude as well — the replies you get here are written by an AI, not typed by a person.

We are telling you this for two reasons. You should know what you are trusting with your notes: read the code, it is MIT and all of it is public. And when a reply is confidently wrong — it happens — say so in the thread. That correction is the mechanism.

Your notes live in a folder. SoloMD is the editor on top — with a first-class agent surface inside the editor, and the MCP endpoint Claude Code / Cursor can drive from outside. Same .md files. Chat with your vault. Schedule recipes that run when you're not at the keyboard. Hand the same vault to any MCP client.

Built on Tauri 2 + Vue 3 + CodeMirror 6. Universal macOS dmg ~32 MB. Free, MIT, no subscription, no SoloMD-hosted servers. Your notes, AI keys, embeddings index, and git history all stay on your machine.

New in 4.6 — the knowledge-graph layer

The structural half Obsidian and Tolaria had and SoloMD didn't — now built, and on SoloMD's own warm design system:

  • Properties inspector (⌘⇧I) — edit YAML frontmatter with type/date/status/relation pickers; line-surgical writes preserve your comments, inline arrays and quoting byte-for-byte.
  • Type-driven sidebar — give a note type: Project and its members become a first-class collapsible section with custom icon/colour/columns (types-as-lenses).
  • Typed relationships — belongs_to / related_to / has frontmatter with automatic inverses, resolved server-side for big vaults.
  • Relationship graph ("Neighborhood") — per-note explorer of outgoing + inverse links; click to navigate, ⌘-click to pivot.
  • Saved filtered views — persistent sidebar views (.solomd/views/*.yml) with a recursive all/any filter builder (type / status / date / regex).
  • Inbox workflow — ⌘E mark-organized + auto-advance to inbox-zero.
  • tldraw whiteboards — Markdown-backed boards (```tldraw fence), cross-compatible with Tolaria's on-disk format, lazy-loaded so the core stays light.

Plus a unified macOS title bar (traffic lights inline in the toolbar) and a from-scratch design-system + Vue component library. Still ~15–32 MB, still free, still local-first. → Compare vs Obsidian / Typora / Tolaria

Three halves of one product

The editor. WYSIWYG live edit (Typora-style), tabs + split panes, KaTeX + Mermaid, image paste to _assets/, slideshow mode (⌘⌥P), Vim mode, Hunspell + CJK proofread, semantic search (⌘⇧F), wikilinks + backlinks, Pandoc export. CJK encodings (GBK / Big5 / Shift-JIS) auto-detected.

The endpoint. A bundled solomd-mcp binary exposes the same vault to any MCP client — 16 tools out of the box, including SoloMD-only ones (autogit_log, autogit_diff, autogit_rollback, sync_status, share_url, read_agent_trace) that no other markdown server has. v4.0 adds --workspace path1 --workspace path2 federation — one MCP session, many vaults. Plus a solomd agent <prompt> CLI that hands off to Claude Code / Codex CLI with the MCP pre-wired.

The agent surface (v4.0). Right-side Agent Panel: streamed chat-with-vault, [[wikilink]] citations, tool-call cards inline, Insert / Copy buttons drop the reply into the active note. Plus declarative recipes as YAML in <workspace>/.solomd/agents/*.yml — cron / on-save / on-commit / on-tag-add / manual triggers. Every agent write lands on its own AutoGit branch you accept or reject before it touches main; write-cap default 5; refuses to start when the working tree is dirty; replayable trace.jsonl per run with read_agent_trace MCP tool.

| Feature | | |---|---| | Agent Panel (v4.0) | Streamed chat-with-vault peer to Outline / Backlinks / Tags / History. Tool-call cards expand inline; reply Insert / Copy to active editor; run history persists as plain markdown under .solomd/agent-runs/. | | Scheduled recipes (v4.0) | YAML jobs in your vault. AutoGit branch sandbox + accept/reject UI before merge. Per-run write-cap (default 5, ceiling 50). 11-recipe cookbook ships in-tree. | | Replayable trace (v4.0) | trace.jsonl per step (prompt / model_call / tool_call / tool_result / git_commit). Replay-from-step rewinds and re-executes. | | Workspace federation (v4.0) | solomd-mcp --workspace path1 --workspace path2. One Claude Desktop session, many vaults. MCP profiles UI in Settings → Integrations. | | Ollama first-class (v4.0) | Auto-detect at localhost:11434. Three model presets (qwen2.5:1.5b/7b/14b). provider: local recipe alias for zero-cloud-cost autonomous loops. | | AI rewrite, BYOK | 14 providers — OpenAI · Claude · Gemini · DeepSeek · Qwen · GLM · Kimi · Doubao · SiliconFlow · OpenRouter · Mistral · Groq · xAI · Ollama. Direct vendor calls. Keys in OS keychain. | | GitHub-backed sync | Push your vault to a private GitHub repo on every save. Optional E2EE (Argon2id + XChaCha20-Poly1305). GitLab / Gitea / any HTTPS git URL works too. | | AutoGit per note | Every ⌘S is a commit in a local .git inside the workspace. libgit2 vendored, no system git needed. Never auto-pushed. | | MCP server bundled | solomd-mcp ships in the install. 16 tools (see the list). stdio by default; opt-in Streamable HTTP (--transport http, loopback + bearer token). Read-only by default; --allow-write opt-in. | | REST API (v4.0) | Localhost only, token auth. Same surface as MCP for clients that don't speak MCP yet — Alfred / Raycast / n8n / your own scripts. | | BYOK cost meter (v4.0) | Per-provider running tokens-spent counter, opt-in. Settings → Integrations. | | Cloud-folder mode | If your vault lives in ~/Library/Mobile Documents/... or ~/Dropbox/..., SoloMD detects it and adds cross-device session restore on top — the OS already does the file sync. | | Public read-only share | Command palette → copy solomd.app/share/?repo=...&path=... link. Renders any file in your public GitHub repo, no SoloMD account needed to view. |

Use it

After installing SoloMD on macOS / Linux:

1. Chat with your vault. Open the right-side Agent Panel (⌘⇧P → "View: Toggle Agent Panel" if hidden). Streamed multi-turn against your notes; tool-call cards show every read/write inline. Reply too long? Insert drops it into the active note at the cursor (replaces selection); Copy to clipboard.

2. Schedule a recipe. Settings → Recipes → Browse cookbook. 11 starters ready: weekly review, daily summary, TODO extraction, translation pass, citation cleanup, CJK proofread agent, link-rot detector, frontmatter normalizer, outline-to-blog, refactor pass, weekly tag triage. Install one, edit the prompt, run it.

3. Drive the same vault from another LLM client. One-shot (needs the solomd CLI):

# Print the MCP config snippet for your AI client…
solomd mcp-config
# …or write it straight into every AI client found on this machine.
solomd mcp install
{
  "mcpServers": {
    "solomd": {
      "command": "/Applications/SoloMD.app/Contents/MacOS/solomd-mcp",
      "args": ["--workspace", "/Users/me/Documents/SoloMD"]
    }
  }
}

Paste into Claude Desktop / Cursor / etc. For multi-vault federation, repeat --workspace:

"args": [
  "--workspace", "/Users/me/Documents/SoloMD",
  "--workspace", "/Users/me/Documents/work-notes"
]

4. Or hand a prompt to claude / codex CLI directly:

solomd agent "rewrite this week of dailies into a weekly review and commit it"

Path-traversal guarded. The LLM only sees what you point the workspace at.

MCP tools

| Kind | Tools | |---|---| | Read (always on) | list_notes, read_note, search, get_backlinks, get_outline, list_tags, list_tasks, autogit_log, autogit_diff, sync_status, share_url, read_agent_trace | | Write (--allow-write) | write_note, append_to_note, autogit_rollback, export_note |

export_note (html / md / txt / docx) runs the headless exporter with Node.js, so it is only listed when a SoloMD source checkout with pnpm install is available (or SOLOMD_EXPORT_SCRIPT points at its app/scripts/solomd-export.mjs); it never overwrites an existing file unless the call passes overwrite: true. Line numbers in every tool count from the top of the file, front matter included.

Transports: stdio (default, what Claude Desktop / Cursor / Cline / Codex use), or Streamable HTTP for hosted clients:

solomd-mcp --workspace ~/Documents/SoloMD --transport http --bind 127.0.0.1:8765 --auth-token "$TOKEN"

HTTP binds to loopback by default; with --auth-token (or SOLOMD_MCP_TOKEN) every request needs Authorization: Bearer <token>. See mcp-server/README.md.

Command line (CLI)

solomd is a small bash CLI (macOS / Linux) for scripting the vault and wiring up MCP:

curl -fsSL https://raw.githubusercontent.com/zhitongblog/solomd/main/scripts/install-cli.sh | bash
solomd help

open, new, list, search, cat, agent, mcp list|install|uninstall|config, mcp-config, --version; every command takes -h. The notes folder is $SOLOMD_NOTES (default ~/Documents/SoloMD). solomd export uses the same Node exporter as export_note, so it needs a source checkout: run scripts/solomd from a clone with pnpm install done in app/, or set SOLOMD_EXPORT_SCRIPT. In the app itself, use File → Export.

Install

Latest release and every download: the releases page.

System requirements: Windows 10+, macOS 10.15+, current mainstream Linux, iOS 15+, Android 7+ (API 24). Windows 7/8/8.1 cannot be supported — the Rust toolchain requires Windows 10 (since Rust 1.78) and Microsoft froze WebView2 (SoloMD's rendering engine) at version 109 on Windows 7 with no security updates — and no legacy build exists.

macOS — universal dmg (Apple Silicon + Intel, signed + notarized)

brew install

Truncated for display — read the full file on GitHub.

Related Skills

View on GitHub
GitHub Stars1.4k
CategoryAI
Updated5h ago
Forks97

Languages

TypeScript

Trust signals

100/100

From repository metadata: license, adoption, age and documentation. Not a code audit — see the Safety scan above for what the skill file itself contains.

No cautions