diff --git a/ChangeLog.d/00README.md b/ChangeLog.d/00README.md
new file mode 100644
index 0000000..b559e23
--- /dev/null
+++ b/ChangeLog.d/00README.md
@@ -0,0 +1,67 @@
+# Pending changelog entry directory
+
+This directory contains changelog entries that have not yet been merged
+to the changelog file ([`../ChangeLog`](../ChangeLog)).
+
+## Changelog entry file format
+
+A changelog entry file must have the extension `*.txt` and must have the
+following format:
+
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+Security
+   * Change description.
+   * Another change description.
+
+Features
+   * Yet another change description. This is a long change description that
+     spans multiple lines.
+   * Yet again another change description.
+
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+The permitted changelog entry categories are as follows:
+<!-- Keep this synchronized with STANDARD_CATEGORIES in assemble_changelog.py! -->
+
+    API changes
+    Default behavior changes
+    Requirement changes
+    New deprecations
+    Removals
+    Features
+    Security
+    Bugfix
+    Changes
+
+Use “Changes” for anything that doesn't fit in the other categories, such as
+performance, documentation and test improvements.
+
+## How to write a changelog entry
+
+Each entry starts with three spaces, an asterisk and a space. Continuation
+lines start with 5 spaces. Lines wrap at 79 characters.
+
+Write full English sentences with proper capitalization and punctuation. Use
+the present tense. Use the imperative where applicable. For example: “Fix a
+bug in mbedtls_xxx() ….”
+
+Include GitHub issue numbers where relevant. Use the format “#1234” for an
+Mbed TLS issue. Add other external references such as CVE numbers where
+applicable.
+
+Credit the author of the contribution if the contribution is not a member of
+the Mbed TLS development team. Also credit bug reporters where applicable.
+
+**Explain why, not how**. Remember that the audience is the users of the
+library, not its developers. In particular, for a bug fix, explain the
+consequences of the bug, not how the bug was fixed. For a new feature, explain
+why one might be interested in the feature. For an API change or a deprecation,
+explain how to update existing applications.
+
+See [existing entries](../ChangeLog) for examples.
+
+## How `ChangeLog` is updated
+
+Run [`../scripts/assemble_changelog.py`](../scripts/assemble_changelog.py)
+from a Git working copy
+to move the entries from files in `ChangeLog.d` to the main `ChangeLog` file.
diff --git a/scripts/assemble_changelog.py b/scripts/assemble_changelog.py
new file mode 100755
index 0000000..ffa3f16
--- /dev/null
+++ b/scripts/assemble_changelog.py
@@ -0,0 +1,505 @@
+#!/usr/bin/env python3
+
+"""Assemble Mbed TLS change log entries into the change log file.
+
+Add changelog entries to the first level-2 section.
+Create a new level-2 section for unreleased changes if needed.
+Remove the input files unless --keep-entries is specified.
+
+In each level-3 section, entries are sorted in chronological order
+(oldest first). From oldest to newest:
+* Merged entry files are sorted according to their merge date (date of
+  the merge commit that brought the commit that created the file into
+  the target branch).
+* Committed but unmerged entry files are sorted according to the date
+  of the commit that adds them.
+* Uncommitted entry files are sorted according to their modification time.
+
+You must run this program from within a git working directory.
+"""
+
+# Copyright (C) 2019, Arm Limited, All Rights Reserved
+# SPDX-License-Identifier: Apache-2.0
+#
+# Licensed under the Apache License, Version 2.0 (the "License"); you may
+# not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
+# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+#
+# This file is part of Mbed TLS (https://tls.mbed.org)
+
+import argparse
+from collections import OrderedDict, namedtuple
+import datetime
+import functools
+import glob
+import os
+import re
+import subprocess
+import sys
+
+class InputFormatError(Exception):
+    def __init__(self, filename, line_number, message, *args, **kwargs):
+        message = '{}:{}: {}'.format(filename, line_number,
+                                     message.format(*args, **kwargs))
+        super().__init__(message)
+
+class CategoryParseError(Exception):
+    def __init__(self, line_offset, error_message):
+        self.line_offset = line_offset
+        self.error_message = error_message
+        super().__init__('{}: {}'.format(line_offset, error_message))
+
+class LostContent(Exception):
+    def __init__(self, filename, line):
+        message = ('Lost content from {}: "{}"'.format(filename, line))
+        super().__init__(message)
+
+# The category names we use in the changelog.
+# If you edit this, update ChangeLog.d/README.md.
+STANDARD_CATEGORIES = (
+    b'API changes',
+    b'Default behavior changes',
+    b'Requirement changes',
+    b'New deprecations',
+    b'Removals',
+    b'Features',
+    b'Security',
+    b'Bugfix',
+    b'Changes',
+)
+
+CategoryContent = namedtuple('CategoryContent', [
+    'name', 'title_line', # Title text and line number of the title
+    'body', 'body_line', # Body text and starting line number of the body
+])
+
+class ChangelogFormat:
+    """Virtual class documenting how to write a changelog format class."""
+
+    @classmethod
+    def extract_top_version(cls, changelog_file_content):
+        """Split out the top version section.
+
+        If the top version is already released, create a new top
+        version section for an unreleased version.
+
+        Return ``(header, top_version_title, top_version_body, trailer)``
+        where the "top version" is the existing top version section if it's
+        for unreleased changes, and a newly created section otherwise.
+        To assemble the changelog after modifying top_version_body,
+        concatenate the four pieces.
+        """
+        raise NotImplementedError
+
+    @classmethod
+    def version_title_text(cls, version_title):
+        """Return the text of a formatted version section title."""
+        raise NotImplementedError
+
+    @classmethod
+    def split_categories(cls, version_body):
+        """Split a changelog version section body into categories.
+
+        Return a list of `CategoryContent` the name is category title
+        without any formatting.
+        """
+        raise NotImplementedError
+
+    @classmethod
+    def format_category(cls, title, body):
+        """Construct the text of a category section from its title and body."""
+        raise NotImplementedError
+
+class TextChangelogFormat(ChangelogFormat):
+    """The traditional Mbed TLS changelog format."""
+
+    _unreleased_version_text = b'= mbed TLS x.x.x branch released xxxx-xx-xx'
+    @classmethod
+    def is_released_version(cls, title):
+        # Look for an incomplete release date
+        return not re.search(br'[0-9x]{4}-[0-9x]{2}-[0-9x]?x', title)
+
+    _top_version_re = re.compile(br'(?:\A|\n)(=[^\n]*\n+)(.*?\n)(?:=|$)',
+                                 re.DOTALL)
+    @classmethod
+    def extract_top_version(cls, changelog_file_content):
+        """A version section starts with a line starting with '='."""
+        m = re.search(cls._top_version_re, changelog_file_content)
+        top_version_start = m.start(1)
+        top_version_end = m.end(2)
+        top_version_title = m.group(1)
+        top_version_body = m.group(2)
+        if cls.is_released_version(top_version_title):
+            top_version_end = top_version_start
+            top_version_title = cls._unreleased_version_text + b'\n\n'
+            top_version_body = b''
+        return (changelog_file_content[:top_version_start],
+                top_version_title, top_version_body,
+                changelog_file_content[top_version_end:])
+
+    @classmethod
+    def version_title_text(cls, version_title):
+        return re.sub(br'\n.*', version_title, re.DOTALL)
+
+    _category_title_re = re.compile(br'(^\w.*)\n+', re.MULTILINE)
+    @classmethod
+    def split_categories(cls, version_body):
+        """A category title is a line with the title in column 0."""
+        if not version_body:
+            return []
+        title_matches = list(re.finditer(cls._category_title_re, version_body))
+        if not title_matches or title_matches[0].start() != 0:
+            # There is junk before the first category.
+            raise CategoryParseError(0, 'Junk found where category expected')
+        title_starts = [m.start(1) for m in title_matches]
+        body_starts = [m.end(0) for m in title_matches]
+        body_ends = title_starts[1:] + [len(version_body)]
+        bodies = [version_body[body_start:body_end].rstrip(b'\n') + b'\n'
+                  for (body_start, body_end) in zip(body_starts, body_ends)]
+        title_lines = [version_body[:pos].count(b'\n') for pos in title_starts]
+        body_lines = [version_body[:pos].count(b'\n') for pos in body_starts]
+        return [CategoryContent(title_match.group(1), title_line,
+                                body, body_line)
+                for title_match, title_line, body, body_line
+                in zip(title_matches, title_lines, bodies, body_lines)]
+
+    @classmethod
+    def format_category(cls, title, body):
+        # `split_categories` ensures that each body ends with a newline.
+        # Make sure that there is additionally a blank line between categories.
+        if not body.endswith(b'\n\n'):
+            body += b'\n'
+        return title + b'\n' + body
+
+class ChangeLog:
+    """An Mbed TLS changelog.
+
+    A changelog file consists of some header text followed by one or
+    more version sections. The version sections are in reverse
+    chronological order. Each version section consists of a title and a body.
+
+    The body of a version section consists of zero or more category
+    subsections. Each category subsection consists of a title and a body.
+
+    A changelog entry file has the same format as the body of a version section.
+
+    A `ChangelogFormat` object defines the concrete syntax of the changelog.
+    Entry files must have the same format as the changelog file.
+    """
+
+    # Only accept dotted version numbers (e.g. "3.1", not "3").
+    # Refuse ".x" in a version number where x is a letter: this indicates
+    # a version that is not yet released. Something like "3.1a" is accepted.
+    _version_number_re = re.compile(br'[0-9]+\.[0-9A-Za-z.]+')
+    _incomplete_version_number_re = re.compile(br'.*\.[A-Za-z]')
+
+    def add_categories_from_text(self, filename, line_offset,
+                                 text, allow_unknown_category):
+        """Parse a version section or entry file."""
+        try:
+            categories = self.format.split_categories(text)
+        except CategoryParseError as e:
+            raise InputFormatError(filename, line_offset + e.line_offset,
+                                   e.error_message)
+        for category in categories:
+            if not allow_unknown_category and \
+               category.name not in self.categories:
+                raise InputFormatError(filename,
+                                       line_offset + category.title_line,
+                                       'Unknown category: "{}"',
+                                       category.name.decode('utf8'))
+            self.categories[category.name] += category.body
+
+    def __init__(self, input_stream, changelog_format):
+        """Create a changelog object.
+
+        Populate the changelog object from the content of the file
+        input_stream.
+        """
+        self.format = changelog_format
+        whole_file = input_stream.read()
+        (self.header,
+         self.top_version_title, top_version_body,
+         self.trailer) = self.format.extract_top_version(whole_file)
+        # Split the top version section into categories.
+        self.categories = OrderedDict()
+        for category in STANDARD_CATEGORIES:
+            self.categories[category] = b''
+        offset = (self.header + self.top_version_title).count(b'\n') + 1
+        self.add_categories_from_text(input_stream.name, offset,
+                                      top_version_body, True)
+
+    def add_file(self, input_stream):
+        """Add changelog entries from a file.
+        """
+        self.add_categories_from_text(input_stream.name, 1,
+                                      input_stream.read(), False)
+
+    def write(self, filename):
+        """Write the changelog to the specified file.
+        """
+        with open(filename, 'wb') as out:
+            out.write(self.header)
+            out.write(self.top_version_title)
+            for title, body in self.categories.items():
+                if not body:
+                    continue
+                out.write(self.format.format_category(title, body))
+            out.write(self.trailer)
+
+
+@functools.total_ordering
+class EntryFileSortKey:
+    """This classes defines an ordering on changelog entry files: older < newer.
+
+    * Merged entry files are sorted according to their merge date (date of
+      the merge commit that brought the commit that created the file into
+      the target branch).
+    * Committed but unmerged entry files are sorted according to the date
+      of the commit that adds them.
+    * Uncommitted entry files are sorted according to their modification time.
+
+    This class assumes that the file is in a git working directory with
+    the target branch checked out.
+    """
+
+    # Categories of files. A lower number is considered older.
+    MERGED = 0
+    COMMITTED = 1
+    LOCAL = 2
+
+    @staticmethod
+    def creation_hash(filename):
+        """Return the git commit id at which the given file was created.
+
+        Return None if the file was never checked into git.
+        """
+        hashes = subprocess.check_output(['git', 'log', '--format=%H',
+                                          '--follow',
+                                          '--', filename])
+        m = re.search(b'(.+)$', hashes)
+        if not m:
+            # The git output is empty. This means that the file was
+            # never checked in.
+            return None
+        # The last commit in the log is the oldest one, which is when the
+        # file was created.
+        return m.group(0)
+
+    @staticmethod
+    def list_merges(some_hash, target, *options):
+        """List merge commits from some_hash to target.
+
+        Pass options to git to select which commits are included.
+        """
+        text = subprocess.check_output(['git', 'rev-list',
+                                        '--merges', *options,
+                                        b'..'.join([some_hash, target])])
+        return text.rstrip(b'\n').split(b'\n')
+
+    @classmethod
+    def merge_hash(cls, some_hash):
+        """Return the git commit id at which the given commit was merged.
+
+        Return None if the given commit was never merged.
+        """
+        target = b'HEAD'
+        # List the merges from some_hash to the target in two ways.
+        # The ancestry list is the ones that are both descendants of
+        # some_hash and ancestors of the target.
+        ancestry = frozenset(cls.list_merges(some_hash, target,
+                                             '--ancestry-path'))
+        # The first_parents list only contains merges that are directly
+        # on the target branch. We want it in reverse order (oldest first).
+        first_parents = cls.list_merges(some_hash, target,
+                                        '--first-parent', '--reverse')
+        # Look for the oldest merge commit that's both on the direct path
+        # and directly on the target branch. That's the place where some_hash
+        # was merged on the target branch. See
+        # https://stackoverflow.com/questions/8475448/find-merge-commit-which-include-a-specific-commit
+        for commit in first_parents:
+            if commit in ancestry:
+                return commit
+        return None
+
+    @staticmethod
+    def commit_timestamp(commit_id):
+        """Return the timestamp of the given commit."""
+        text = subprocess.check_output(['git', 'show', '-s',
+                                        '--format=%ct',
+                                        commit_id])
+        return datetime.datetime.utcfromtimestamp(int(text))
+
+    @staticmethod
+    def file_timestamp(filename):
+        """Return the modification timestamp of the given file."""
+        mtime = os.stat(filename).st_mtime
+        return datetime.datetime.fromtimestamp(mtime)
+
+    def __init__(self, filename):
+        """Determine position of the file in the changelog entry order.
+
+        This constructor returns an object that can be used with comparison
+        operators, with `sort` and `sorted`, etc. Older entries are sorted
+        before newer entries.
+        """
+        self.filename = filename
+        creation_hash = self.creation_hash(filename)
+        if not creation_hash:
+            self.category = self.LOCAL
+            self.datetime = self.file_timestamp(filename)
+            return
+        merge_hash = self.merge_hash(creation_hash)
+        if not merge_hash:
+            self.category = self.COMMITTED
+            self.datetime = self.commit_timestamp(creation_hash)
+            return
+        self.category = self.MERGED
+        self.datetime = self.commit_timestamp(merge_hash)
+
+    def sort_key(self):
+        """"Return a concrete sort key for this entry file sort key object.
+
+        ``ts1 < ts2`` is implemented as ``ts1.sort_key() < ts2.sort_key()``.
+        """
+        return (self.category, self.datetime, self.filename)
+
+    def __eq__(self, other):
+        return self.sort_key() == other.sort_key()
+
+    def __lt__(self, other):
+        return self.sort_key() < other.sort_key()
+
+
+def check_output(generated_output_file, main_input_file, merged_files):
+    """Make sanity checks on the generated output.
+
+    The intent of these sanity checks is to have reasonable confidence
+    that no content has been lost.
+
+    The sanity check is that every line that is present in an input file
+    is also present in an output file. This is not perfect but good enough
+    for now.
+    """
+    generated_output = set(open(generated_output_file, 'rb'))
+    for line in open(main_input_file, 'rb'):
+        if line not in generated_output:
+            raise LostContent('original file', line)
+    for merged_file in merged_files:
+        for line in open(merged_file, 'rb'):
+            if line not in generated_output:
+                raise LostContent(merged_file, line)
+
+def finish_output(changelog, output_file, input_file, merged_files):
+    """Write the changelog to the output file.
+
+    The input file and the list of merged files are used only for sanity
+    checks on the output.
+    """
+    if os.path.exists(output_file) and not os.path.isfile(output_file):
+        # The output is a non-regular file (e.g. pipe). Write to it directly.
+        output_temp = output_file
+    else:
+        # The output is a regular file. Write to a temporary file,
+        # then move it into place atomically.
+        output_temp = output_file + '.tmp'
+    changelog.write(output_temp)
+    check_output(output_temp, input_file, merged_files)
+    if output_temp != output_file:
+        os.rename(output_temp, output_file)
+
+def remove_merged_entries(files_to_remove):
+    for filename in files_to_remove:
+        os.remove(filename)
+
+def list_files_to_merge(options):
+    """List the entry files to merge, oldest first.
+
+    "Oldest" is defined by `EntryFileSortKey`.
+    """
+    files_to_merge = glob.glob(os.path.join(options.dir, '*.txt'))
+    files_to_merge.sort(key=EntryFileSortKey)
+    return files_to_merge
+
+def merge_entries(options):
+    """Merge changelog entries into the changelog file.
+
+    Read the changelog file from options.input.
+    Read entries to merge from the directory options.dir.
+    Write the new changelog to options.output.
+    Remove the merged entries if options.keep_entries is false.
+    """
+    with open(options.input, 'rb') as input_file:
+        changelog = ChangeLog(input_file, TextChangelogFormat)
+    files_to_merge = list_files_to_merge(options)
+    if not files_to_merge:
+        sys.stderr.write('There are no pending changelog entries.\n')
+        return
+    for filename in files_to_merge:
+        with open(filename, 'rb') as input_file:
+            changelog.add_file(input_file)
+    finish_output(changelog, options.output, options.input, files_to_merge)
+    if not options.keep_entries:
+        remove_merged_entries(files_to_merge)
+
+def show_file_timestamps(options):
+    """List the files to merge and their timestamp.
+
+    This is only intended for debugging purposes.
+    """
+    files = list_files_to_merge(options)
+    for filename in files:
+        ts = EntryFileSortKey(filename)
+        print(ts.category, ts.datetime, filename)
+
+def set_defaults(options):
+    """Add default values for missing options."""
+    output_file = getattr(options, 'output', None)
+    if output_file is None:
+        options.output = options.input
+    if getattr(options, 'keep_entries', None) is None:
+        options.keep_entries = (output_file is not None)
+
+def main():
+    """Command line entry point."""
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument('--dir', '-d', metavar='DIR',
+                        default='ChangeLog.d',
+                        help='Directory to read entries from'
+                             ' (default: ChangeLog.d)')
+    parser.add_argument('--input', '-i', metavar='FILE',
+                        default='ChangeLog',
+                        help='Existing changelog file to read from and augment'
+                             ' (default: ChangeLog)')
+    parser.add_argument('--keep-entries',
+                        action='store_true', dest='keep_entries', default=None,
+                        help='Keep the files containing entries'
+                             ' (default: remove them if --output/-o is not specified)')
+    parser.add_argument('--no-keep-entries',
+                        action='store_false', dest='keep_entries',
+                        help='Remove the files containing entries after they are merged'
+                             ' (default: remove them if --output/-o is not specified)')
+    parser.add_argument('--output', '-o', metavar='FILE',
+                        help='Output changelog file'
+                             ' (default: overwrite the input)')
+    parser.add_argument('--list-files-only',
+                        action='store_true',
+                        help=('Only list the files that would be processed '
+                              '(with some debugging information)'))
+    options = parser.parse_args()
+    set_defaults(options)
+    if options.list_files_only:
+        show_file_timestamps(options)
+        return
+    merge_entries(options)
+
+if __name__ == '__main__':
+    main()