diff --git a/src/main.zig b/src/main.zig
index b6761e1..1512820 100644
--- a/src/main.zig
+++ b/src/main.zig
@@ -13,19 +13,8 @@ pub fn main() !void {
     x_auth_key = std.os.getenv("CF_X_AUTH_KEY").?;
     // TODO: All this stuff needs to be different
     //
-    // TODO: We need to break index.js into the wrangler-generated bundling thing
-    //       and the actual code we're using to run the wasm file.
-    //       We might actually want a "run this wasm" upload vs a "these are my
-    //       js files" upload. But for now we'll optimize for wasm
-    const index = @embedFile("index.js");
-    const wasm = @embedFile("demo.wasm");
-    const memfs = @embedFile("dist/memfs.wasm");
     const accountid = @embedFile("accountid.txt");
     const worker_name = @embedFile("worker_name.txt");
-    const deploy_request = @embedFile("deploy_request.txt");
-    _ = deploy_request;
-    const put_script = "https://api.cloudflare.com/client/v4/accounts/{s}/workers/scripts/{s}?include_subdomain_availability=true&excludeScript=true";
-    _ = put_script;
     const get_subdomain = "https://api.cloudflare.com/client/v4/accounts/{s}/workers/scripts/{s}/subdomain";
     _ = get_subdomain;
 
@@ -36,19 +25,90 @@ pub fn main() !void {
     var bw = std.io.bufferedWriter(stdout_file);
     const stdout = bw.writer();
 
-    try stdout.print("Index bytes: {d}\n", .{index.len});
-    try stdout.print("Wasm bytes: {d}\n", .{wasm.len});
-    try stdout.print("Memfs bytes: {d}\n", .{memfs.len});
     try stdout.print("Account: {s}\n", .{accountid});
     try stdout.print("Worker name: {s}\n", .{worker_name});
 
+    // Determine if worker exists. This lets us know if we need to enable it later
     const worker_exists = try workerExists(allocator, &client, accountid, worker_name);
     try stdout.print(
         "Worker exists: {}\n",
         .{worker_exists},
     );
+
+    try putNewWorker(allocator, &client, accountid, worker_name);
     try bw.flush(); // don't forget to flush!
 }
+fn putNewWorker(allocator: std.mem.Allocator, client: *std.http.Client, account_id: []const u8, name: []const u8) !void {
+    const put_script = "https://api.cloudflare.com/client/v4/accounts/{s}/workers/scripts/{s}?include_subdomain_availability=true&excludeScript=true";
+    const url = try std.fmt.allocPrint(allocator, put_script, .{ account_id, name });
+    defer allocator.free(url);
+    // TODO: All this stuff needs to be different
+    //
+    // TODO: We need to break index.js into the wrangler-generated bundling thing
+    //       and the actual code we're using to run the wasm file.
+    //       We might actually want a "run this wasm" upload vs a "these are my
+    //       js files" upload. But for now we'll optimize for wasm
+    const script = @embedFile("index.js");
+    const wasm = @embedFile("demo.wasm");
+    const memfs = @embedFile("dist/memfs.wasm");
+    const deploy_request =
+        "------formdata-undici-032998177938\r\n" ++
+        "Content-Disposition: form-data; name=\"metadata\"\r\n\r\n" ++
+        "{{\"main_module\":\"index.js\",\"bindings\":[],\"compatibility_date\":\"2023-10-02\",\"compatibility_flags\":[]}}\r\n" ++
+        "------formdata-undici-032998177938\r\n" ++
+        "Content-Disposition: form-data; name=\"index.js\"; filename=\"index.js\"\r\n" ++
+        "Content-Type: application/javascript+module\r\n" ++
+        "\r\n" ++
+        "{[script]s}\r\n" ++
+        "------formdata-undici-032998177938\r\n" ++
+        "Content-Disposition: form-data; name=\"./24526702f6c3ed7fb02b15125f614dd38804525f-demo.wasm\"; filename=\"./24526702f6c3ed7fb02b15125f614dd38804525f-demo.wasm\"\r\n" ++
+        "Content-Type: application/wasm\r\n" ++
+        "\r\n" ++
+        "{[wasm]s}\r\n" ++
+        "------formdata-undici-032998177938\r\n" ++
+        "Content-Disposition: form-data; name=\"./c5f1acc97ad09df861eff9ef567c2186d4e38de3-memfs.wasm\"; filename=\"./c5f1acc97ad09df861eff9ef567c2186d4e38de3-memfs.wasm\"\r\n" ++
+        "Content-Type: application/wasm\r\n" ++
+        "\r\n" ++
+        "{[memfs]s}\r\n" ++
+        "------formdata-undici-032998177938--";
+
+    var headers = std.http.Headers.init(allocator);
+    defer headers.deinit();
+    try headers.append("X-Auth-Email", x_auth_email);
+    try headers.append("X-Auth-Key", x_auth_key);
+    // TODO: fix this
+    try headers.append("Content-Type", "multipart/form-data; boundary=----formdata-undici-032998177938");
+    const request_payload = try std.fmt.allocPrint(allocator, deploy_request, .{
+        .script = script,
+        .wasm = wasm,
+        .memfs = memfs,
+    });
+    defer allocator.free(request_payload);
+    // Get content length. For some reason it's forcing a chunked transfer type without this.
+    // That's not entirely a bad idea, but for now I want to match what I see
+    // coming through wrangler
+    const cl = try std.fmt.allocPrint(allocator, "{d}", .{request_payload.len});
+    defer allocator.free(cl);
+    try headers.append("Content-Length", cl);
+    std.debug.print("payload length {d}\n", .{request_payload.len});
+    const blah1 = "------formdata-undici-032998177938\r\nC";
+    std.debug.print("first 80 bytes {s}", .{request_payload[0..80]});
+    std.debug.print("Starts with? {}", .{std.mem.startsWith(u8, request_payload, blah1)});
+    var req = try client.request(.PUT, try std.Uri.parse(url), headers, .{});
+    defer req.deinit();
+
+    req.transfer_encoding = .{ .content_length = @as(u64, request_payload.len) };
+    try req.start();
+    try req.writeAll(request_payload);
+    try req.finish();
+    try req.wait();
+    // std.debug.print("Status is {}\n", .{req.response.status});
+    // std.debug.print("Url is {s}\n", .{url});
+    if (req.response.status != .ok) {
+        std.debug.print("Status is {}\n", .{req.response.status});
+        return error.RequestFailed;
+    }
+}
 
 fn workerExists(allocator: std.mem.Allocator, client: *std.http.Client, account_id: []const u8, name: []const u8) !bool {
     const existence_check = "https://api.cloudflare.com/client/v4/accounts/{s}/workers/services/{s}";