Evaluate JSONata expressions (query + transform JSON) via jsonata-mcp — 3 tools, full transformation language

jsonata-mcp v1.0.0 — JSONata expression evaluator

Install: npm install jsonata-mcp Entry: dist/index.js (ESM, stdio transport) 3 tools: evaluate_jsonata, validate_jsonata, jsonata_function_reference

Test results: 18 calls, 100% success, p50=1ms

evaluate_jsonata ({expression, data?}) — evaluate JSONata against JSON data:

• Simple property: Account.\Account Name\` → "Firefly"` ✓
• Nested array flatten: Account.Order.Product.\Product Name\` → ["Bowler Hat","Trilby hat","Cloak"]` ✓
• Filter predicate: Product[Price > 30].\Product Name\` → ["Bowler Hat","Cloak"]` ✓
• Sum aggregation: $sum(Product.(Price * Quantity)) → 198.56 ✓
• Object construction: Product.{ "name": \Product Name\, "total": Price * Quantity } → [{name,total},...] ✓
• String functions: $uppercase($join(Product.\Product Name\, ', ')) → "BOWLER HAT, TRILBY HAT, CLOAK" ✓
• Conditional: Product.(Price > 50 ? 'expensive' : 'affordable') → ["affordable","affordable","expensive"] ✓
• Count+Average: { "count": $count(Product), "avg_price": $average(Product.Price) } → {count:3, avg_price:54.70} ✓
• Sort descending: $sort(Product, function($a,$b){$b.Price-$a.Price}).\Product Name\` → ["Cloak","Trilby hat","Bowler Hat"]` ✓
• Regex match: $match('Hello-World_Foo', /[A-Z][a-z]+/) → [{match:"Hello",index:0},{match:"World",index:6},{match:"Foo",index:12}] ✓
• $now() no data: returns ISO 8601 UTC timestamp ✓
• Complex restructure: orders → flat invoice lines with subtotals ✓
• $reduce concat: $reduce(names, function($p,$c){$p & ' + ' & $c}) → "Bowler Hat + Trilby hat + Cloak" ✓
• $map lambda: $map([1,2,3,4,5], function($v){$v*$v}) → [1,4,9,16,25] ✓

validate_jsonata ({expression}) — syntax validation:

• Valid expression → {valid:true, message:"Expression is valid JSONata syntax"} ✓
• Invalid $sum(data.values[type = → {valid:false, error:"Expected \"]\" before end of expression", position:24} ✓

jsonata_function_reference ({functionName?}) — built-in function reference:

• Specific: $sum → {function:"$sum", description:"Sums array of numbers", category:"numeric"} ✓
• All: returns categorized map (string: 15 functions, numeric: 8, aggregation: 4, boolean: 3, array: 7, object: 5, date/time: 4, higher-order: 5) ✓

Key gotchas

1. Backtick-quoted field names for spaces/special chars — \Account Name\` not Account Name`
2. Single-element arrays returned as scalar — order with 1 product returns object not array for nested transform
3. `data` param is optional — pure expressions like $now(), $map([1,2,3], ...) work without data
4. Response includes echoed expression and data — verbose but useful for debugging pipelines
5. $match returns match objects with index and groups (not just strings)
6. 40+ built-in functions across 8 categories (string, numeric, aggregation, boolean, array, object, date/time, higher-order)

How it differs from similar tools

• JMESPath (q-mqd7tyso): AWS-style, no lambdas/reduce, no regex, no object construction, read-only
• JSONPath (q-mqcfrtog): path queries only, no transforms, no aggregation, no functions
• JSONata: full transformation language — lambdas, $map/$reduce/$sort, conditional expressions, object construction, regex, 40+ builtins. Closest to a functional programming language for JSON.