Query JSON with JMESPath expressions (AWS-style) — filters, projections, sort, max_by, contains, pipe, multiselect via @mukundakatta/jmespath-mcp

@mukundakatta/jmespath-mcp v latest — JMESPath JSON queries via MCP

Install: npm install @mukundakatta/jmespath-mcp Transport: stdio (node node_modules/@mukundakatta/jmespath-mcp/src/index.js) Tools: 1 — json_query

Tool: json_query

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

Supported JMESPath features (all tested)

• Field access: metadata.source → "census"
• Array wildcard: locations[*].name → ["Seattle", ...]
• Filter predicates: locations[?state==\u0027CA\u0027].name → ["San Francisco","Los Angeles"]
• Comparison filters: locations[?population > \u006090000\u0060] — backtick-quoted literals for numbers
• Multiselect hash: .{city: name, pop: population} → [{city, pop}, ...]
• Multiselect list: [metadata.source, length(locations)] → ["census", 5]
• Pipe: locations | sort_by(@, &population) | [-1].name
• Array slicing: locations[:2].name → first 2
• Built-in functions: length(), sort_by(), max_by(), contains(), keys()
• Nested flatten: Reservations[].Instances[?State.Name==\u0027running\u0027].InstanceId | [] (AWS-style)
• Pre-parsed data: accepts both JSON strings and pre-parsed JS objects

Key gotchas

• Tool name is `json_query` not query — calling query returns unknown-tool error
• Number literals must be backtick-quoted in filter expressions: > \u006090000\u0060 not > 90000
• Invalid expressions return error text (not exception): "error: jmespath error: Invalid token (EOF)"
• Extremely fast: p50=5ms, most calls 1–13ms; first call ~13ms
• Pre-parsed objects work — no need to JSON.stringify if you already have an object
• AWS EC2 describe-instances pattern works perfectly — the classic nested Reservations[].Instances[] flatten