Markra vs Muesli

Markra is a local-first, open-source Markdown workspace with native AI review, offered as a desktop app and web editor. It supports WYSIWYG and source editing, with inline rendering for links, images, tables, tasks, callouts, code, math, and Mermaid diagrams. Users can open files or folders and keep local files, tabs, and rich editing together; documents remain ordinary Markdown files, and writing does not require an account. AI can answer questions, rewrite or summarize text, and propose edits, but users must approve changes before they are applied. AI features require at least one configured provider, including cloud providers, Ollama, or OpenAI-compatible endpoints. Provider API keys and custom headers are stored in local settings, and exported settings can include them. Notes can optionally sync through WebDAV, which preserves conflict copies. Desktop downloads are available for macOS, Windows, and Linux. Desktop exports include HTML, PDF, and Pandoc formats when configured; the web editor supports HTML download and browser print or PDF. Core features are free; AI use requires a configured provider.

Muesli is a free, real-time collaborative Markdown editor for teams and AI agents working in the same plain .md file. Collaborators can see live cursors and presence, discuss material in comment threads, submit tracked suggestions, and use append-only history with snapshots. Editing tools include inline preview, a source toggle, KaTeX math, Mermaid diagrams, editable tables and callouts, wikilinks, and backlinks. Claude Code, Cursor, Codex, Copilot, and Gemini CLI can work through MCP or edit the file directly. When a human is present in a document, agent edits become suggestions for review. Guest links can grant viewer, commenter, or editor access without requiring accounts. Storage choices include local files, S3-compatible services, GitHub, Gitea, Forgejo, and Google Drive. Muesli is available on the web, Windows, macOS, Linux, and as a self-hosted service, and is open source under AGPL-3.0. Self-hosting uses one Docker image, but durable documents and history require Postgres and Redis. The hosted service says document contents remain in user-chosen storage and are not read, mined, or used for training.