diff --git a/.gitignore b/.gitignore
index 9986ac6b1..1092d097a 100644
--- a/.gitignore
+++ b/.gitignore
@@ -61,6 +61,7 @@ llama-batched-swift
 /rpc-server
 out/
 tmp/
+autogen-*.md
 
 # Deprecated
 
diff --git a/Makefile b/Makefile
index 9c61d3ec0..ba3f11c53 100644
--- a/Makefile
+++ b/Makefile
@@ -39,6 +39,7 @@ BUILD_TARGETS = \
 	llama-tokenize \
 	llama-vdot \
 	llama-cvector-generator \
+	llama-export-docs \
 	tests/test-c.o
 
 # Binaries only useful for tests
@@ -1449,6 +1450,12 @@ examples/server/%.hpp: examples/server/public/% Makefile
 		echo "unsigned int $${NAME}_len = $(shell cat $< | wc -c );" \
 	) > $@
 
+llama-export-docs: examples/export-docs/export-docs.cpp \
+	$(OBJ_ALL)
+	$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
+	$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
+	./llama-export-docs
+
 libllava.a: examples/llava/llava.cpp \
 	examples/llava/llava.h \
 	examples/llava/clip.cpp \
diff --git a/common/common.cpp b/common/common.cpp
index 49db551ae..2d99bfc25 100644
--- a/common/common.cpp
+++ b/common/common.cpp
@@ -425,7 +425,7 @@ bool gpt_params_parse_ex(int argc, char ** argv, gpt_params & params, std::vecto
             throw std::invalid_argument(format(
                 "error while handling argument \"%s\": %s\n\n"
                 "usage:\n%s\n\nto show complete usage, run with -h",
-                arg.c_str(), e.what(), arg_to_options[arg]->to_string(false).c_str()));
+                arg.c_str(), e.what(), arg_to_options[arg]->to_string().c_str()));
         }
     }
 
@@ -582,14 +582,13 @@ static std::vector<std::string> break_str_into_lines(std::string input, size_t m
     return result;
 }
 
-std::string llama_arg::to_string(bool markdown) {
+std::string llama_arg::to_string() {
     // params for printing to console
     const static int n_leading_spaces = 40;
     const static int n_char_per_line_help = 70; // TODO: detect this based on current console
     std::string leading_spaces(n_leading_spaces, ' ');
 
     std::ostringstream ss;
-    if (markdown) ss << "| `";
     for (const auto & arg : args) {
         if (arg == args.front()) {
             ss << (args.size() == 1 ? arg : format("%-7s", (arg + ",").c_str()));
@@ -598,20 +597,16 @@ std::string llama_arg::to_string(bool markdown) {
         }
     }
     if (!value_hint.empty()) ss << " " << value_hint;
-    if (!markdown) {
-        if (ss.tellp() > n_leading_spaces - 3) {
-            // current line is too long, add new line
-            ss << "\n" << leading_spaces;
-        } else {
-            // padding between arg and help, same line
-            ss << std::string(leading_spaces.size() - ss.tellp(), ' ');
-        }
-        const auto help_lines = break_str_into_lines(help, n_char_per_line_help);
-        for (const auto & line : help_lines) {
-            ss << (&line == &help_lines.front() ? "" : leading_spaces) << line << "\n";
-        }
+    if (ss.tellp() > n_leading_spaces - 3) {
+        // current line is too long, add new line
+        ss << "\n" << leading_spaces;
     } else {
-        ss << "` | " << help << " |";
+        // padding between arg and help, same line
+        ss << std::string(leading_spaces.size() - ss.tellp(), ' ');
+    }
+    const auto help_lines = break_str_into_lines(help, n_char_per_line_help);
+    for (const auto & line : help_lines) {
+        ss << (&line == &help_lines.front() ? "" : leading_spaces) << line << "\n";
     }
     return ss.str();
 }
@@ -619,7 +614,7 @@ std::string llama_arg::to_string(bool markdown) {
 void gpt_params_print_usage(std::vector<llama_arg> & options) {
     auto print_options = [](std::vector<llama_arg *> & options) {
         for (llama_arg * opt : options) {
-            printf("%s", opt->to_string(false).c_str());
+            printf("%s", opt->to_string().c_str());
         }
     };
 
diff --git a/common/common.h b/common/common.h
index f84948330..7536120fc 100644
--- a/common/common.h
+++ b/common/common.h
@@ -349,7 +349,7 @@ struct llama_arg {
         return std::getenv(env.c_str());
     }
 
-    std::string to_string(bool markdown);
+    std::string to_string();
 };
 
 std::vector<llama_arg> gpt_params_parser_init(gpt_params & params, llama_example ex);
diff --git a/examples/export-docs/CMakeLists.txt b/examples/export-docs/CMakeLists.txt
new file mode 100644
index 000000000..0e953167e
--- /dev/null
+++ b/examples/export-docs/CMakeLists.txt
@@ -0,0 +1,5 @@
+set(TARGET llama-export-docs)
+add_executable(${TARGET} export-docs.cpp)
+install(TARGETS ${TARGET} RUNTIME)
+target_link_libraries(${TARGET} PRIVATE common llama ${CMAKE_THREAD_LIBS_INIT})
+target_compile_features(${TARGET} PRIVATE cxx_std_11)
diff --git a/examples/export-docs/export-docs.cpp b/examples/export-docs/export-docs.cpp
new file mode 100644
index 000000000..e21c4b89d
--- /dev/null
+++ b/examples/export-docs/export-docs.cpp
@@ -0,0 +1,47 @@
+#include "common.h"
+#include "llama.h"
+
+#include <cmath>
+#include <cstdio>
+#include <fstream>
+#include <string>
+#include <vector>
+
+// Export usage message (-h) to markdown format
+
+static void export_md(std::string fname, llama_example ex) {
+    std::ofstream file(fname, std::ofstream::out | std::ofstream::trunc);
+
+    gpt_params params;
+    auto options = gpt_params_parser_init(params, ex);
+
+    file << "| Argument | Explanation |\n";
+    file << "| -------- | ----------- |\n";
+    for (auto & opt : options) {
+        file << "| `";
+        // args
+        for (const auto & arg : opt.args) {
+        if (arg == opt.args.front()) {
+                file << (opt.args.size() == 1 ? arg : (arg + ", "));
+            } else {
+                file << arg << (arg != opt.args.back() ? ", " : "");
+            }
+        }
+        // value hint
+        std::string md_value_hint(opt.value_hint);
+        string_replace_all(md_value_hint, "|", "\\|");
+        file << " " << md_value_hint;
+        // help text
+        std::string md_help(opt.help);
+        string_replace_all(md_help, "\n", "<br/>");
+        string_replace_all(md_help, "|", "\\|");
+        file << "` | " << md_help << " |\n";
+    }
+}
+
+int main(int, char **) {
+    export_md("autogen-main.md", LLAMA_EXAMPLE_MAIN);
+    export_md("autogen-server.md", LLAMA_EXAMPLE_SERVER);
+
+    return 0;
+}