jsonask

Query JSON with natural language instead of jq syntax. Stop memorizing jq — just ask.

Usage

# Pipe JSON and ask a question
cat data.json | jsonask "show names"

# Pass file path + query
jsonask data.json "count by status"

# Extract specific fields
jsonask users.json "extract all emails"

# Filter data
jsonask users.json "where role is admin"
jsonask products.json "price greater than 100"

# Sort
jsonask products.json "sort by price"
jsonask products.json "sort by price desc"

# Aggregation
jsonask data.json "count by status"
jsonask data.json "unique categories"

# Limit results
jsonask data.json "top 5 by score"
jsonask data.json "first 3"

# Explore structure
jsonask data.json "keys"

Output Formats

# Table (default)
jsonask data.json "show names"

# JSON
jsonask data.json "show names" -o json

# List (values only)
jsonask data.json "show names" -o list

How It Works

jsonask parses natural language queries into structured operations using pattern matching. It handles:

Operation	Example Queries
Extract	”show names”, “get emails”, “extract all ids”
Filter	”where status is active”, “price greater than 100”
Count	”count by role”, “how many items”
Unique	”unique values”, “distinct categories”
Sort	”sort by name”, “top 5 by score”
Explore	”keys”, “what fields”

Features

• Zero config — works immediately, no API key needed
• Offline — local query engine handles common patterns
• Pipe-friendly — reads from stdin or file
• Smart parsing — fuzzy key matching, dot notation support
• Multiple outputs — table, JSON, or raw list format

Practical JSON Query Examples

These are the queries that come up most often. Every one runs locally with no API key.

Explore an unknown payload. The first thing to do with unfamiliar JSON is learn its shape:

jsonask api-response.json "keys"
jsonask api-response.json "what fields"

Extract one field across an array. Instead of remembering .[] | .name, ask:

jsonask users.json "show names"
jsonask users.json "extract all emails"
jsonask orders.json "get ids"

Filter by value. Equality and numeric comparison both work:

jsonask users.json "where role is admin"
jsonask users.json "where status is active"
jsonask products.json "price greater than 100"

Aggregate. Counts and distinct values are one query:

jsonask events.json "count by type"
jsonask events.json "how many items"
jsonask catalog.json "unique categories"

Sort and limit. Ordering combined with a top-N is common in reports:

jsonask products.json "sort by price desc"
jsonask players.json "top 5 by score"
jsonask results.json "first 3"

Use dot notation for nested keys. Fuzzy matching means close spellings still resolve:

jsonask config.json "show server.port"
jsonask data.json "extract user.email"

Chain with other tools. Because jsonask reads stdin and writes JSON with -o json, it composes:

curl -s https://api.example.com/items \
  | jsonask "where status is active" -o json \
  | jq '.[0]'

jsonask vs. Alternatives

Tool	Query input	Needs learning	Offline	Best for
jsonask	Natural language	No	Yes	Quick, ad-hoc queries without jq syntax
jq	A dedicated DSL	Yes, steep	Yes	Complex, reproducible transformations
fx	JavaScript expressions	Some JS	Yes	Interactive exploration with full JS
jless	Interactive viewer	Minimal	Yes	Browsing large JSON by hand

The honest recommendation: keep jq for scripts and pipelines you commit, because its syntax is precise and reproducible. Use jsonask for the exploratory 80% — when you just want the answer now and the query will not be saved.

FAQ

Does jsonask call an LLM? No. It uses local pattern matching to translate common phrasings into operations. That is why it needs no API key and works offline.

What happens with a query it does not understand? It reports that it could not parse the query. Try the keys or what fields query first to confirm the field names. Paraphrase using the operations in the table above.

Can it modify JSON or write files? No. jsonask is read-only: it extracts, filters, aggregates, sorts, and prints.

Which output format should I use? Table for reading, -o json for piping into another tool, -o list for a clean list of scalar values.

Is it a full jq replacement? No. It covers common extract/filter/aggregate/sort queries. For arbitrary transformation logic in a committed script, use jq.

Related Tools

• DomainForge — produces the RAG datasets you then query with jsonask.
• secretsweep — inspect JSON configs before they are committed.
• gitcommitgen — commit the query changes you settle on.
• Browse the rest of the developer tools collection and the collection announcement.