diff --git a/bin/jpp b/bin/jpp
new file mode 100755
index 00000000..965fa64e
--- /dev/null
+++ b/bin/jpp
@@ -0,0 +1,7 @@
+#!/usr/bin/env python
+
+import sys
+from jmespath.jpp import jpp_main
+
+if __name__ == "__main__":
+    sys.exit(jpp_main(argv=sys.argv))
diff --git a/jmespath/jpp.py b/jmespath/jpp.py
new file mode 100644
index 00000000..c2cefb7a
--- /dev/null
+++ b/jmespath/jpp.py
@@ -0,0 +1,291 @@
+import argparse
+import json
+import os
+import pprint
+import sys
+import itertools
+
+import jmespath.exceptions
+
+# This 2 space indent matches https://github.com/jmespath/jp behavior.
+JP_COMPAT_DUMP_KWARGS = (
+    ("indent", 2),
+    ("ensure_ascii", False),
+)
+
+
+def decode_json_stream(stream):
+    """
+    Decode a text JSON input stream and generate objects until EOF.
+    """
+    eof = False
+    line_buffer = []
+    while line_buffer or not eof:
+        progress = False
+        if not eof:
+            line = stream.readline()
+            if line:
+                progress = True
+                line_buffer.append(line)
+            else:
+                eof = True
+
+        if line_buffer:
+            chunk = "".join(line_buffer)
+            del line_buffer[:]
+
+            try:
+                yield json.loads(chunk)
+                progress = True
+            except json.JSONDecodeError as e:
+                if e.pos > 0:
+                    try:
+                        yield json.loads(chunk[: e.pos])
+                        progress = True
+                    except json.JSONDecodeError:
+                        # Raise if there's no progress, since a given
+                        # chunk should be growning if it is not yet
+                        # decodable.
+                        if not progress:
+                            raise
+                        line_buffer.append(chunk)
+                    else:
+                        line_buffer.append(chunk[e.pos :])
+                else:
+                    raise
+
+
+def jpp_main(argv=None):
+    argv = sys.argv if argv is None else argv
+    parser = argparse.ArgumentParser(
+        prog=os.path.basename(argv[0]),
+    )
+    parser.add_argument("expression", nargs="?", default=None)
+    parser.add_argument(
+        "-a",
+        "--accumulate",
+        action="store_true",
+        dest="accumulate",
+        default=False,
+        help=(
+            "Accumulate all output objects into a single recursively merged output object."
+        ),
+    )
+    parser.add_argument(
+        "-c",
+        "--compact",
+        action="store_true",
+        dest="compact",
+        default=False,
+        help=("Produce compact JSON output that omits nonessential whitespace."),
+    )
+    parser.add_argument(
+        "-e",
+        "--expr-file",
+        dest="expr_file",
+        default=None,
+        help=("Read JMESPath expression from the specified file."),
+    )
+    parser.add_argument(
+        "-f",
+        "--filename",
+        dest="filename",
+        default=None,
+        help=(
+            "The filename containing the input data. "
+            "If a filename is not given then data is "
+            "read from stdin."
+        ),
+    )
+    parser.add_argument(
+        "-r",
+        "--raw-output",
+        action="store_false",
+        dest="quoted",
+        default=True,
+        help=(
+            "If the final result is a string, it will be printed without quotes (an alias for --unquoted)."
+        ),
+    )
+    parser.add_argument(
+        "-R",
+        "--raw-input",
+        action="store_true",
+        dest="raw_input",
+        default=False,
+        help=("Read raw string input and box it as JSON strings."),
+    )
+    parser.add_argument(
+        "-s",
+        "--slurp",
+        action="store_true",
+        dest="slurp",
+        default=False,
+        help=(
+            "Read one or more input JSON objects into an array and apply the JMESPath expression to the resulting array."
+        ),
+    )
+    parser.add_argument(
+        "-u",
+        "--unbox",
+        action="store_true",
+        dest="unbox",
+        default=False,
+        help=(
+            "If the final result is a list, unbox it into a stream of output objects that is suitable for consumption by --slurp mode."
+        ),
+    )
+    parser.add_argument(
+        "--unquoted",
+        action="store_false",
+        dest="quoted",
+        default=True,
+        help=("If the final result is a string, it will be printed without quotes."),
+    )
+    parser.add_argument(
+        "--ast",
+        action="store_true",
+        help=(
+            "Only print the AST of the parsed expression.  Do not rely on this output, only useful for debugging purposes."
+        ),
+    )
+    parser.usage = "{}\n  {}".format(
+        parser.format_usage().partition("usage: ")[-1],
+        "jpp is an extended superset of the jp CLI for JMESPath",
+    )
+
+    args = parser.parse_args(argv[1:])
+    expression = args.expression
+    if expression == "help":
+        parser.print_help()
+        return 1
+
+    if expression and args.expr_file:
+        parser.error("Only use one of the expression or --expr-file arguments.")
+
+    dump_kwargs = dict(JP_COMPAT_DUMP_KWARGS)
+    if args.compact:
+        dump_kwargs.pop("indent", None)
+        dump_kwargs["separators"] = (",", ":")
+
+    if args.expr_file:
+        with open(args.expr_file, "rt") as f:
+            expression = f.read()
+        if not expression:
+            raise jmespath.exceptions.EmptyExpressionError()
+    elif not expression:
+        expression = "@"
+
+    if args.ast:
+        # Only print the AST
+        expression = jmespath.compile(args.expression)
+        sys.stdout.write(pprint.pformat(expression.parsed))
+        sys.stdout.write("\n")
+        return 0
+
+    if args.filename:
+        f = open(args.filename, "rt")
+    else:
+        f = sys.stdin
+
+    accumulator = None
+    stream_iter = None
+    eof = False
+
+    with f:
+        if not args.raw_input:
+            stream_iter = decode_json_stream(f)
+        while True:
+            while True:
+                if args.slurp:
+                    if stream_iter is None:
+                        data = f.read()
+                    else:
+                        data = list(stream_iter)
+                elif stream_iter is None:
+                    data = f.readline()
+                else:
+                    try:
+                        data = next(stream_iter)
+                    except StopIteration:
+                        data = None
+
+                if not data:
+                    eof = True
+                    break
+
+                result = jmespath.search(expression, data)
+
+                if args.accumulate:
+                    if accumulator is None:
+                        accumulator = result
+                    else:
+                        accumulator = merge(accumulator, result)
+                    if args.slurp:
+                        break
+                else:
+                    break
+
+            if args.accumulate:
+                result = accumulator
+            elif eof:
+                break
+
+            if args.unbox and isinstance(result, list):
+                for element in result:
+                    output_result(args, dump_kwargs, element)
+            else:
+                output_result(args, dump_kwargs, result)
+
+            if eof or args.accumulate or args.slurp:
+                break
+    return 0
+
+
+def output_result(args, dump_kwargs, result):
+    if args.quoted or not isinstance(result, str):
+        result = json.dumps(result, **dump_kwargs)
+
+    sys.stdout.write(result)
+
+    if args.quoted or (
+        not args.quoted and isinstance(result, str) and result[-1:] != "\n"
+    ):
+        sys.stdout.write("\n")
+
+
+def merge(base, head):
+    """
+    Recursively merge head onto base.
+    """
+    if isinstance(head, dict):
+        if not isinstance(base, dict):
+            return head
+
+        result = {}
+        for k in itertools.chain(head, base):
+            try:
+                result[k] = merge(base[k], head[k])
+            except KeyError:
+                try:
+                    result[k] = head[k]
+                except KeyError:
+                    result[k] = base[k]
+
+    elif isinstance(head, list):
+        result = []
+        if isinstance(base, list):
+            result.extend(base)
+            for node in head:
+                if node not in result:
+                    result.append(node)
+        else:
+            result.extend(head)
+    else:
+        result = head
+
+    return result
+
+
+if __name__ == "__main__":
+    sys.exit(jpp_main(argv=sys.argv))
diff --git a/setup.py b/setup.py
index 3337d517..637b70e7 100644
--- a/setup.py
+++ b/setup.py
@@ -24,7 +24,7 @@
     author='James Saryerwinnie',
     author_email='js@jamesls.com',
     url='https://github.com/jmespath/jmespath.py',
-    scripts=['bin/jp.py'],
+    scripts=['bin/jp.py', 'bin/jpp'],
     packages=find_packages(exclude=['tests']),
     license='MIT',
     python_requires='>=2.6, !=3.0.*, !=3.1.*, !=3.2.*',