diff --git a/templates/tests/test_unicode_decimals.html b/templates/tests/test_unicode_decimals.html
new file mode 100644
index 0000000000..caa7a27f1f
--- /dev/null
+++ b/templates/tests/test_unicode_decimals.html
@@ -0,0 +1,5 @@
+header
+
+{{ render_markdown_path("zerver/tests/markdown/test_unicode_decimals.md", {"unescape_rendered_html": unescape_rendered_html}) }}
+
+footer
diff --git a/templates/zerver/tests/markdown/test_unicode_decimals.md b/templates/zerver/tests/markdown/test_unicode_decimals.md
new file mode 100644
index 0000000000..c60c140460
--- /dev/null
+++ b/templates/zerver/tests/markdown/test_unicode_decimals.md
@@ -0,0 +1 @@
+{}
diff --git a/zerver/templatetags/app_filters.py b/zerver/templatetags/app_filters.py
index 0d356592dc..ddda2b0bf5 100644
--- a/zerver/templatetags/app_filters.py
+++ b/zerver/templatetags/app_filters.py
@@ -1,4 +1,5 @@
 import os
+from html import unescape
 from typing import Any, Dict, List, Optional
 
 import markdown
@@ -141,4 +142,13 @@ def render_markdown_path(markdown_file_path: str,
     html = md_engine.convert(markdown_string)
     rendered_html = jinja.from_string(html).render(context)
 
+    if context.get('unescape_rendered_html', False):
+        # In some exceptional cases (such as our Freshdesk webhook docs),
+        # code blocks in some of our Markdown templates have characters such
+        # as '{' encoded as '{' to prevent clashes with Jinja2 syntax,
+        # but the encoded form never gets decoded because the text ends up
+        # inside a <pre> tag. So here, we explicitly "unescape" such characters
+        # if 'unescape_rendered_html' is True.
+        rendered_html = unescape(rendered_html)
+
     return mark_safe(rendered_html)
diff --git a/zerver/tests/test_templates.py b/zerver/tests/test_templates.py
index 4e5d7e80a9..0272d4d782 100644
--- a/zerver/tests/test_templates.py
+++ b/zerver/tests/test_templates.py
@@ -216,6 +216,22 @@ class TemplateTestCase(ZulipTestCase):
         self.assertEqual(content_sans_whitespace,
                          'header<h1id="hello">Hello!</h1><p>Thisissome<em>boldtext</em>.</p>footer')
 
+    def test_encoded_unicode_decimals_in_markdown_template(self) -> None:
+        template = get_template("tests/test_unicode_decimals.html")
+        context = {'unescape_rendered_html': False}
+        content = template.render(context)
+
+        content_sans_whitespace = content.replace(" ", "").replace('\n', '')
+        self.assertEqual(content_sans_whitespace,
+                         'header<p>&#123;&#125;</p>footer')
+
+        context = {'unescape_rendered_html': True}
+        content = template.render(context)
+
+        content_sans_whitespace = content.replace(" ", "").replace('\n', '')
+        self.assertEqual(content_sans_whitespace,
+                         'header<p>{}</p>footer')
+
     def test_markdown_nested_code_blocks(self) -> None:
         template = get_template("tests/test_markdown.html")
         context = {
diff --git a/zerver/views/integrations.py b/zerver/views/integrations.py
index e614bc4aa6..64a6a2325f 100644
--- a/zerver/views/integrations.py
+++ b/zerver/views/integrations.py
@@ -146,6 +146,14 @@ def integration_doc(request: HttpRequest, integration_name: str=REQ(default=None
         context['hubot_docs_url'] = integration.hubot_docs_url
     if isinstance(integration, EmailIntegration):
         context['email_gateway_example'] = settings.EMAIL_GATEWAY_EXAMPLE
+    if integration.name == 'freshdesk':
+        # In our Freshdesk docs, some nested code blocks have characters such
+        # as '{' encoded as '&#123;' to prevent clashes with Jinja2 syntax,
+        # but the encoded form never gets rendered because the text ends up
+        # inside a <pre> tag. So here, we explicitly set a directive that
+        # a particular template should be "unescaped" before being displayed.
+        # Note that this value is used by render_markdown_path.
+        context['unescape_rendered_html'] = True
 
     doc_html_str = render_markdown_path(integration.doc, context)