diff --git a/chibicc.h b/chibicc.h
index 5b8c759f78daeaee5682df5a66df6ec7b685013f..6f99ac44ec1bee3cfdc91f04a7428ab8754fca86 100644
--- a/chibicc.h
+++ b/chibicc.h
@@ -72,11 +72,18 @@ void error_tok(Token *tok, char *fmt, ...);
 bool equal(Token *tok, char *op);
 Token *skip(Token *tok, char *op);
 bool consume(Token **rest, Token *tok, char *str);
+void convert_keywords(Token *tok);
 Token *tokenize_file(char *filename);
 
 #define unreachable() \
   error("internal error at %s:%d", __FILE__, __LINE__)
 
+//
+// preprocess.c
+//
+
+Token *preprocess(Token *tok);
+
 //
 // parse.c
 //
diff --git a/main.c b/main.c
index 43c7df29744db2daa0b2bf84eee31f873a3cc0d0..e3cc90e97fefd9ed2d84d23ba624a59b3e05e900 100644
--- a/main.c
+++ b/main.c
@@ -168,6 +168,7 @@ static void run_cc1(int argc, char **argv, char *input, char *output) {
 static void cc1(void) {
   // Tokenize and parse.
   Token *tok = tokenize_file(base_file);
+  tok = preprocess(tok);
   Obj *prog = parse(tok);
 
   // Traverse the AST to emit assembly.
diff --git a/preprocess.c b/preprocess.c
new file mode 100644
index 0000000000000000000000000000000000000000..e26e4d9e9fc0b7553c239613108fd3be88f9799a
--- /dev/null
+++ b/preprocess.c
@@ -0,0 +1,7 @@
+#include "chibicc.h"
+
+// Entry point function of the preprocessor.
+Token *preprocess(Token *tok) {
+  convert_keywords(tok);
+  return tok;
+}
diff --git a/tokenize.c b/tokenize.c
index 76a5c27a3451224022c9a2685dc592e861d6a208..59b79493e0b553eb57dbe16de38dce946490d9f9 100644
--- a/tokenize.c
+++ b/tokenize.c
@@ -350,7 +350,7 @@ static Token *read_number(char *start) {
   return tok;
 }
 
-static void convert_keywords(Token *tok) {
+void convert_keywords(Token *tok) {
   for (Token *t = tok; t->kind != TK_EOF; t = t->next)
     if (is_keyword(t))
       t->kind = TK_KEYWORD;
@@ -446,7 +446,6 @@ static Token *tokenize(char *filename, char *p) {
 
   cur = cur->next = new_token(TK_EOF, p, p);
   add_line_numbers(head.next);
-  convert_keywords(head.next);
   return head.next;
 }