Query JSON data with JMESPath expressions (AWS-style: filters, projections, sort_by, pipes) via @mukundakatta/jmespath-mcp (npx)

Verified Recipe: Query JSON with JMESPath via @mukundakatta/jmespath-mcp

Server: @mukundakatta/jmespath-mcp v latest Transport: stdio via npx @mukundakatta/jmespath-mcp Auth: none (credential-free) Tools: 1 — json_query

Tool: json_query

Params:

• expression (string, required) — JMESPath expression
• data (any, required) — JSON to query. Pass as a JSON string OR a pre-parsed object/array

Returns: the raw JMESPath query result (string, number, array, object, or null)

JMESPath Features Verified

• Dot notation: company -> scalar value
• Wildcard projection: employees[*].name -> array of names
• Filter predicate: employees[?dept == 'engineering'].name -> filtered array
• Numeric filter + multiselect hash: employees[?salary > backtick100000backtick].{name: name, salary: salary}
• Pipe + function: employees[?dept == 'engineering'] | length(@) -> 3
• Flatten nested arrays: employees[*].skills[] -> flat array
• sortby: `sortby(employees, &salary)[*].name` -> sorted
• reverse: reverse(sort_by(employees, &salary)) -> descending
• maxby: `maxby(employees, &salary).name` -> 'Dave'
• OR filter: locations[?state == 'WA' || state == 'OR']
• contains: employees[?contains(skills, 'sql')].name
• Multiselect list: [e[0].name, e[0].dept, e[0].salary]
• JSON string data: pass data as string instead of object — works
• Nested index: employees[0].skills[0] -> 'python'

All 14 features verified with real execution.

Key Observations

• Full JMESPath spec — the same query language used by AWS CLI --query flag
• Data accepts string or object — both work identically
• Numeric literals use backticks in JMESPath expressions
• Multiselect hash {name: name, salary: salary} reshapes output into custom objects
• First call ~80ms (JIT); subsequent calls 2-22ms
• Returns raw values — scalars not wrapped in an object

Gotchas

• No error details on bad expressions — terse error messages
• No streaming — entire data must fit in memory
• String equality is case-sensitive
• Backtick literals required for numbers in filters (without backticks, compares as string)