diff --git a/src/helpgen.zig b/src/helpgen.zig
new file mode 100644
index 000000000..4447346f4
--- /dev/null
+++ b/src/helpgen.zig
@@ -0,0 +1,167 @@
+//! This program is used to generate the help strings from the configuration
+//! file and CLI actions for Ghostty. These can then be used to generate
+//! help, docs, website, etc.
+
+const std = @import("std");
+const ziglyph = @import("ziglyph");
+const Config = @import("config/Config.zig");
+const Action = @import("cli/action.zig").Action;
+
+pub fn main() !void {
+    var gpa = std.heap.GeneralPurposeAllocator(.{}){};
+    const alloc = gpa.allocator();
+
+    const stdout = std.io.getStdOut().writer();
+    try stdout.writeAll(
+        \\// THIS FILE IS AUTO GENERATED
+        \\
+        \\
+    );
+
+    try genConfig(alloc, stdout);
+    try genActions(alloc, stdout);
+}
+
+fn genConfig(alloc: std.mem.Allocator, writer: anytype) !void {
+    var ast = try std.zig.Ast.parse(alloc, @embedFile("config/Config.zig"), .zig);
+    defer ast.deinit(alloc);
+
+    try writer.writeAll(
+        \\/// Configuration help
+        \\pub const Config = struct {
+        \\
+        \\
+    );
+
+    inline for (@typeInfo(Config).Struct.fields) |field| {
+        if (field.name[0] == '_') continue;
+        try genConfigField(alloc, writer, ast, field.name);
+    }
+
+    try writer.writeAll("};\n");
+}
+
+fn genConfigField(
+    alloc: std.mem.Allocator,
+    writer: anytype,
+    ast: std.zig.Ast,
+    comptime field: []const u8,
+) !void {
+    const tokens = ast.tokens.items(.tag);
+    for (tokens, 0..) |token, i| {
+        // We only care about identifiers that are preceded by doc comments.
+        if (token != .identifier) continue;
+        if (tokens[i - 1] != .doc_comment) continue;
+
+        // Identifier may have @"" so we strip that.
+        const name = ast.tokenSlice(@intCast(i));
+        const key = if (name[0] == '@') name[2 .. name.len - 1] else name;
+        if (!std.mem.eql(u8, key, field)) continue;
+
+        const comment = try extractDocComments(alloc, ast, @intCast(i - 1), tokens);
+        try writer.writeAll("pub const ");
+        try writer.writeAll(name);
+        try writer.writeAll(": [:0]const u8 = \n");
+        try writer.writeAll(comment);
+        try writer.writeAll("\n");
+        break;
+    }
+}
+
+fn genActions(alloc: std.mem.Allocator, writer: anytype) !void {
+    try writer.writeAll(
+        \\
+        \\/// Actions help
+        \\pub const Action = struct {
+        \\
+        \\
+    );
+
+    inline for (@typeInfo(Action).Enum.fields) |field| {
+        const action_file = comptime action_file: {
+            const action = @field(Action, field.name);
+            break :action_file action.file();
+        };
+
+        var ast = try std.zig.Ast.parse(alloc, @embedFile(action_file), .zig);
+        defer ast.deinit(alloc);
+        const tokens: []std.zig.Token.Tag = ast.tokens.items(.tag);
+
+        for (tokens, 0..) |token, i| {
+            // We're looking for a function named "run".
+            if (token != .keyword_fn) continue;
+            if (!std.mem.eql(u8, ast.tokenSlice(@intCast(i + 1)), "run")) continue;
+
+            // The function must be preceded by a doc comment.
+            if (tokens[i - 2] != .doc_comment) {
+                std.debug.print(
+                    "doc comment must be present on run function of the {s} action!",
+                    .{field.name},
+                );
+                std.process.exit(1);
+            }
+
+            const comment = try extractDocComments(alloc, ast, @intCast(i - 2), tokens);
+            try writer.writeAll("pub const @\"");
+            try writer.writeAll(field.name);
+            try writer.writeAll("\" = \n");
+            try writer.writeAll(comment);
+            try writer.writeAll("\n\n");
+            break;
+        }
+    }
+
+    try writer.writeAll("};\n");
+}
+
+fn extractDocComments(
+    alloc: std.mem.Allocator,
+    ast: std.zig.Ast,
+    index: std.zig.Ast.TokenIndex,
+    tokens: []std.zig.Token.Tag,
+) ![]const u8 {
+    // Find the first index of the doc comments. The doc comments are
+    // always stacked on top of each other so we can just go backwards.
+    const start_idx: usize = start_idx: for (0..index) |i| {
+        const reverse_i = index - i - 1;
+        const token = tokens[reverse_i];
+        if (token != .doc_comment) break :start_idx reverse_i + 1;
+    } else unreachable;
+
+    // Go through and build up the lines.
+    var lines = std.ArrayList([]const u8).init(alloc);
+    defer lines.deinit();
+    for (start_idx..index + 1) |i| {
+        const token = tokens[i];
+        if (token != .doc_comment) break;
+        try lines.append(ast.tokenSlice(@intCast(i))[3..]);
+    }
+
+    // Convert the lines to a multiline string.
+    var buffer = std.ArrayList(u8).init(alloc);
+    const writer = buffer.writer();
+    const prefix = findCommonPrefix(lines);
+    for (lines.items) |line| {
+        try writer.writeAll("    \\\\");
+        try writer.writeAll(line[@min(prefix, line.len)..]);
+        try writer.writeAll("\n");
+    }
+    try writer.writeAll(";\n");
+
+    return buffer.toOwnedSlice();
+}
+
+fn findCommonPrefix(lines: std.ArrayList([]const u8)) usize {
+    var m: usize = std.math.maxInt(usize);
+    for (lines.items) |line| {
+        var n: usize = std.math.maxInt(usize);
+        for (line, 0..) |c, i| {
+            if (c != ' ') {
+                n = i;
+                break;
+            }
+        }
+        m = @min(m, n);
+    }
+    return m;
+}