mq

name: mq description: "Query markdown content like jq (headings, code blocks, tables, links)"

mq - Markdown Query Language

A jq-like CLI for querying and filtering markdown AST nodes. Repository available at https://github.com/harehare/mq

Core Selectors

Node Attributes

Access with dot notation: selector.attribute

Basic Usage

# Extract all headings
mq '.h' file.md

# H2 headings only
mq '.h2' file.md

# All code blocks
mq '.code' file.md

# All links
mq '.link' file.md

# Tables
mq '.table' file.md

Output Formats

mq '.code' file.md               # markdown (default)
mq -F text '.code' file.md       # plain text - best for code content
mq -F json '.h' file.md          # JSON - best for parsing/analysis
mq -F html '.h' file.md          # HTML output

Filtering with select()

select() is the primary filter mechanism. When the condition uses a node attribute, the selector is implicit:

# Code blocks by language
mq 'select(.code.lang == "typescript")' file.md
mq 'select(.code.lang == "bash")' file.md
mq 'select(.code.lang != "bash")' file.md   # exclude a language

# Headings by depth
mq 'select(.h.depth <= 2)' file.md          # H1 and H2 only
mq 'select(.h.depth == 3)' file.md          # H3 only

# Filter then extract attribute
mq 'select(.code.lang == "python") | .code.value' file.md
mq 'select(.h.depth == 2) | .h.value' file.md

Filtering with contains() and starts_with()

# Heading containing a word
mq '.h | select(contains("Installation"))' file.md

# Link URL containing a domain
mq '.link | select(contains("github"))' file.md

# Code block containing specific content
mq 'select(.code) | select(contains("TODO"))' file.md

# Heading starting with prefix
mq '.h | select(starts_with("How"))' file.md

Combining Selectors

# H1 or H2 (logical or)
mq 'select(or(.h1, .h2))' file.md

# Node is heading AND level 2 (logical and)
mq 'select(and(.h, .h2))' file.md

Extracting Attribute Values

# Get all heading text
mq '.h.value' file.md

# Get all code languages used
mq '.code.lang' file.md

# Get all link URLs
mq '.link.url' file.md

# Get image alt texts
mq '.image.alt' file.md

# Get raw code content (no fences)
mq '.code.value' file.md
# or equivalently:
mq -F text '.code' file.md

Multi-File Processing

# All headings across files
mq '.h' docs/**/*.md

# H1 from all Pi skills
mq '.h1' ~/.pi/agent/skills/*/SKILL.md

# Aggregate all code into single array
mq -A '.code' examples/*.md

String Functions

mq '.h.value | upcase' file.md
mq '.h.value | downcase' file.md
mq '.code.value | replace("old", "new")' file.md
mq '.link.url | starts_with("https")' file.md
mq '.link.url | ends_with(".org")' file.md
mq '.h.value | split(" ")' file.md

Custom Functions with def

# Define reusable filter
mq 'def is_lang(l): .lang == l; | select(.code) | select(is_lang("typescript"))' file.md

In-place Update with -U

Remove or transform nodes in the file:

# Remove all bash code blocks from file
mq -U 'select(.code.lang != "bash")' file.md

# Keep only H1 and H2 headings
mq -U 'select(or(.h1, .h2))' file.md

Practical Pi Agent Patterns

# Get all section titles (H2) from a skill
mq '.h2.value' ~/.pi/agent/skills/python/SKILL.md

# Get all code examples in a specific language
mq 'select(.code.lang == "bash") | .code.value' SKILL.md

# Extract all external links from docs
mq '.link.url' README.md

# Inspect code block structure
mq -F json '.code' file.md

# Check what languages are used in a doc
mq '.code.lang' file.md

# Find headings matching a topic
mq '.h | select(contains("Docker"))' SKILL.md

# Extract YAML frontmatter
mq '.yaml' SKILL.md

Tips

• Use select(.node.attr == val) for filtering — the node selector is implicit in the condition
• Use -F json first to explore node structure when unsure of available attributes
• Use -F text for clean code extraction without fences
• Pipe mq output to standard tools: mq '.code.lang' file.md | sort | uniq -c
• For complex transformations, extract to JSON then pipe to jq