markymark

markymark

A language server and AI agent tool for Markdown and structured data files. Rust workspace with a Zig SIMD parser core. Provides navigation, refactoring, search, and diagnostics across your workspace — as an editor extension (LSP) and AI agent tool (MCP).

Pre-release: APIs and behavior may change before v1.0.

Features

Navigation

• Go to definition — jump to any heading, wiki link, or block ID across files
• Find all references — see every document linking to a heading or symbol
• Rename everywhere — rename headings and update all references in one step

Search

• Symbol search — fuzzy-match headings, tags, and code references (Ctrl+T / Cmd+T)
• Semantic search — rank sections by relevance using embeddings (Voyage API or local ONNX)
• Full-text search — free text with frontmatter, tag, and property filters
• Regex patterns — search file content with context lines and glob filters

Diagnostics

• Broken link detection — wiki links, Markdown links, and heading anchors validated as you type
• Link graph analysis — orphan documents, hub detection, and connected cluster mapping

Formats

• Markdown, JSON, JSONC, JSON5, JSONL, YAML, TOML, .env, INI
• Obsidian + Logseq — wiki links, callouts, block IDs, and page properties

Under the Hood

markymark is a seven-crate Rust workspace with a Zig FFI layer:

• Zig parser core — dependency-free md4c implementation with SIMD-accelerated extraction, statically linked into the final binary
• Cross-document index — string-interned lookup tables with O(1) wiki link resolution, incremental contribution-diffing, and lazy tag maintenance
• Dual protocol — LSP for real-time editing and MCP for AI agent workflows, backed by the same index
• Embedding search — optional semantic index using Voyage API or local ONNX models via fastembed

Install

cargo install markymark-cli

Pre-built binaries: GitHub Releases

• Claude Code: Plugin README

Documentation

Full documentation at sethyanow.github.io/markymark:

• Getting Started
• Editor Setup
• MCP Tools Reference
• Agent Tutorial

Claude Code

Install the plugin for LSP + MCP document intelligence:

claude /plugin install markymark

Add to your CLAUDE.md for LSP-first document reading:

## Document Intelligence
This project uses markymark LSP. Prefer LSP over reading raw files:
- `LSP documentSymbol <file>` — structure/outline before Read
- `LSP hover <file> <line> <col>` — heading backlinks, key path info
- Diagnostics (broken links, duplicate headings) reported automatically
- Works for Markdown, JSON, YAML, TOML, .env, INI, and more

License

AGPL-3.0