backdraw 2.1.0 → 2.1.2

package/README.md

@@ -6,6 +6,11 @@
   <b>Let Claude generate images with your own ChatGPT account, in a background Chrome tab.<br>No API key, no extra cost, nothing to watch.</b>
 </p>
 
+<p align="center">
+  <a href="https://www.npmjs.com/package/backdraw"><img src="https://img.shields.io/npm/v/backdraw" alt="npm version"></a>
+  <a href="https://glama.ai/mcp/servers/devtalh/backdraw"><img src="https://glama.ai/mcp/servers/devtalh/backdraw/badges/score.svg" alt="devtalh/backdraw MCP server"></a>
+</p>
+
 <p align="center">
   <img src="branding/demo/backdraw-demo.gif" alt="Right-click images on the web to save them as references, then Claude Code asks Backdraw for an image in that style; ChatGPT draws it in a background tab and the file is saved on the PC" width="100%">
 </p>
@@ -123,6 +128,8 @@ Run this in PowerShell or cmd (no admin rights needed):
 npx backdraw@latest setup
 ```
 
+It runs the [`backdraw` package from npm](https://www.npmjs.com/package/backdraw).
+
 `setup` does everything except one click in the browser:
 
 1. It copies Backdraw to `%USERPROFILE%\.backdraw` (npx's own copy is temporary) and sets up the
@@ -145,6 +152,28 @@ npx backdraw@latest setup
 The native host is still named `io.github.muhammedtalha04.backdraw`, after the project's old GitHub
 user name: renaming it would break existing installs.
 
+### Install from your client
+
+Your MCP client can also add Backdraw itself; it then starts the server with
+`npx backdraw@latest mcp`. The browser side still needs a one-time setup on the PC, which leaves the
+clients alone:
+
+```
+npx backdraw@latest setup --clients=none
+```
+
+| Client | Add Backdraw |
+|---|---|
+| Claude Code | `/plugin install backdraw --marketplace devtalh/backdraw` (Claude Code 2.1.275 or later; or `/plugin marketplace add devtalh/backdraw`, then `/plugin install backdraw@backdraw`). `/backdraw:setup` explains the browser step. |
+| Cursor | [![Add Backdraw to Cursor](https://cursor.com/deeplink/mcp-install-dark.svg)](https://cursor.com/install-mcp?name=backdraw&config=eyJjb21tYW5kIjoiY21kIiwiYXJncyI6WyIvYyIsIm5weCIsIi15IiwiYmFja2RyYXdAbGF0ZXN0IiwibWNwIl19) |
+| VS Code | [![Install Backdraw in VS Code](https://img.shields.io/badge/VS_Code-Install_Server-0098FF?style=flat-square&logo=visualstudiocode&logoColor=white)](https://insiders.vscode.dev/redirect/mcp/install?name=backdraw&config=%7B%22type%22%3A%22stdio%22%2C%22command%22%3A%22cmd%22%2C%22args%22%3A%5B%22%2Fc%22%2C%22npx%22%2C%22-y%22%2C%22backdraw%40latest%22%2C%22mcp%22%5D%7D) |
+| Codex | `codex mcp add backdraw -- cmd /c npx -y backdraw@latest mcp` |
+| Gemini CLI | `npx backdraw@latest setup --clients=gemini` |
+
+On Windows the command goes through `cmd /c`, because clients start `npx` without a shell. Use one
+way per client: if setup already added Backdraw to Claude Code, the plugin would start a second
+server with the same tools.
+
 ### Manual install
 
 1. **Get the code.** Clone the repository with git, or download it as a ZIP (**Code** →
@@ -388,7 +417,7 @@ themselves stay in your images folder.
 
 | Tool | Parameters (default) | What it does |
 |---|---|---|
-| `generate_image` | `prompt` (required), `aspect_ratio`, `transparent_background` (false), `reference_images`, `reference_collection`, `reference_files`, `conversation_url`, `auto_reattach` (false), `send_as_is` (false), `output_dir`, `wait_seconds` (45), `include_preview` (true) | Starts a new chat in your Project (or continues `conversation_url`) and generates or edits an image. |
+| `generate_image` | `prompt` (required), `aspect_ratio`, `transparent_background` (false), `reference_images`, `reference_collection`, `reference_files`, `conversation_url`, `auto_reattach` (false), `send_as_is` (false), `output_dir`, `filename_hint`, `wait_seconds` (45), `include_preview` (true) | Starts a new chat in your Project (or continues `conversation_url`) and generates or edits an image. |
 | `job_status` | `job_id`, `wait_seconds` (45), `include_preview` (true) | Waits for a job and reports it. Without `job_id`: the bridge status (Chrome connected, selected Project, images folder, queue, recent jobs). |
 | `cancel_job` | `job_id` (required) | Cancels a queued or running job. If ChatGPT is already generating, it presses Stop in the background tab. |
 | `fetch_images` | `conversation_url` (required), `aspect_ratio`, `output_dir`, `wait_seconds` (45), `include_preview` (true) | Downloads the images of the latest answer in an existing chat, without sending anything. |
@@ -406,6 +435,8 @@ themselves stay in your images folder.
 - `send_as_is`: send the prompt word for word, without the short instruction Backdraw adds (e.g. to
   answer a question ChatGPT asked).
 - `output_dir`: also copy the result into this folder; the images folder always keeps a copy.
+- `filename_hint`: a few words naming the subject, used for the file name instead of the start of
+  the prompt (handy when the prompt opens with instructions about the reference images).
 - `wait_seconds`: 0 to 900.
 
 What a result can say:
@@ -435,7 +466,8 @@ waiting; the job keeps running until `cancel_job` stops it.
 
 For setup and troubleshooting, run these in the Backdraw folder (`%USERPROFILE%\.backdraw`, or the
 folder you installed from). From anywhere, `npx backdraw <command>` runs the same commands, e.g.
-`npx backdraw status`. `npm run status` is the same as `node cli.js status`.
+`npx backdraw status`. `npm run status` is the same as `node cli.js status`. `backdraw mcp` runs
+the MCP server itself on stdio; that is what MCP clients start.
 
 ```
 node cli.js status                          Bridge status (Chrome connected, selected project, queue)

package/README.tr.md

@@ -6,6 +6,11 @@
   <b>Claude, kendi ChatGPT hesabınla, Chrome'da arka plandaki bir sekmede görsel üretsin.<br>API anahtarı yok, ek ücret yok, başında beklemene gerek yok.</b>
 </p>
 
+<p align="center">
+  <a href="https://www.npmjs.com/package/backdraw"><img src="https://img.shields.io/npm/v/backdraw" alt="npm sürümü"></a>
+  <a href="https://glama.ai/mcp/servers/devtalh/backdraw"><img src="https://glama.ai/mcp/servers/devtalh/backdraw/badges/score.svg" alt="devtalh/backdraw MCP sunucusu"></a>
+</p>
+
 <p align="center">
   <img src="branding/demo/backdraw-demo.gif" alt="Web'deki görseller sağ tıkla referans olarak kaydediliyor, sonra Claude Code Backdraw'dan o tarzda bir görsel istiyor; ChatGPT görseli arka plan sekmesinde çiziyor ve dosya bilgisayara kaydediliyor" width="100%">
 </p>
@@ -123,6 +128,8 @@ Bunu PowerShell'de ya da cmd'de çalıştır (yönetici izni gerekmez):
 npx backdraw@latest setup
 ```
 
+Bu komut [npm'deki `backdraw` paketini](https://www.npmjs.com/package/backdraw) çalıştırır.
+
 `setup`, tarayıcıdaki tek bir tıklama dışında her şeyi yapar:
 
 1. Backdraw'ı `%USERPROFILE%\.backdraw` klasörüne kopyalar (npx'in kendi kopyası geçicidir) ve görsel
@@ -146,6 +153,28 @@ npx backdraw@latest setup
 Native host'un adı, projenin eski GitHub kullanıcı adından gelen
 `io.github.muhammedtalha04.backdraw` olarak kalıyor: adı değiştirmek mevcut kurulumları bozardı.
 
+### İstemcinden kur
+
+MCP istemcin Backdraw'ı kendisi de ekleyebilir; o zaman sunucuyu `npx backdraw@latest mcp` ile
+başlatır. Tarayıcı tarafı yine de bilgisayarda bir kerelik bir kurulum ister; bu kurulum istemcilere
+dokunmaz:
+
+```
+npx backdraw@latest setup --clients=none
+```
+
+| İstemci | Backdraw'ı ekle |
+|---|---|
+| Claude Code | `/plugin install backdraw --marketplace devtalh/backdraw` (Claude Code 2.1.275 ya da daha yeni; ya da önce `/plugin marketplace add devtalh/backdraw`, sonra `/plugin install backdraw@backdraw`). `/backdraw:setup` tarayıcı adımını anlatır. |
+| Cursor | [![Backdraw'ı Cursor'a ekle](https://cursor.com/deeplink/mcp-install-dark.svg)](https://cursor.com/install-mcp?name=backdraw&config=eyJjb21tYW5kIjoiY21kIiwiYXJncyI6WyIvYyIsIm5weCIsIi15IiwiYmFja2RyYXdAbGF0ZXN0IiwibWNwIl19) |
+| VS Code | [![Backdraw'ı VS Code'a kur](https://img.shields.io/badge/VS_Code-Install_Server-0098FF?style=flat-square&logo=visualstudiocode&logoColor=white)](https://insiders.vscode.dev/redirect/mcp/install?name=backdraw&config=%7B%22type%22%3A%22stdio%22%2C%22command%22%3A%22cmd%22%2C%22args%22%3A%5B%22%2Fc%22%2C%22npx%22%2C%22-y%22%2C%22backdraw%40latest%22%2C%22mcp%22%5D%7D) |
+| Codex | `codex mcp add backdraw -- cmd /c npx -y backdraw@latest mcp` |
+| Gemini CLI | `npx backdraw@latest setup --clients=gemini` |
+
+Windows'ta komut `cmd /c` üzerinden çalışır, çünkü istemciler `npx`'i kabuk olmadan başlatır. Her
+istemci için tek yol kullan: setup Backdraw'ı Claude Code'a zaten eklediyse, plugin aynı araçlarla
+ikinci bir sunucu başlatır.
+
 ### Elle kurulum
 
 1. **Kodu indir.** Depoyu git ile klonla ya da ZIP olarak indir (**Code** → **Download ZIP**) ve
@@ -398,7 +427,7 @@ klasöründe kalır.
 
 | Araç | Parametreler (varsayılan) | Ne yapar |
 |---|---|---|
-| `generate_image` | `prompt` (zorunlu), `aspect_ratio`, `transparent_background` (false), `reference_images`, `reference_collection`, `reference_files`, `conversation_url`, `auto_reattach` (false), `send_as_is` (false), `output_dir`, `wait_seconds` (45), `include_preview` (true) | Projende yeni bir sohbet başlatır (ya da `conversation_url`'deki sohbete devam eder) ve görsel üretir veya düzenler. |
+| `generate_image` | `prompt` (zorunlu), `aspect_ratio`, `transparent_background` (false), `reference_images`, `reference_collection`, `reference_files`, `conversation_url`, `auto_reattach` (false), `send_as_is` (false), `output_dir`, `filename_hint`, `wait_seconds` (45), `include_preview` (true) | Projende yeni bir sohbet başlatır (ya da `conversation_url`'deki sohbete devam eder) ve görsel üretir veya düzenler. |
 | `job_status` | `job_id`, `wait_seconds` (45), `include_preview` (true) | Bir işi bekler ve durumunu bildirir. `job_id` olmadan köprünün durumunu verir (Chrome bağlı mı, seçili Proje, görsel klasörü, kuyruk, son işler). |
 | `cancel_job` | `job_id` (zorunlu) | Kuyruktaki ya da çalışan bir işi iptal eder. ChatGPT üretmeye başladıysa arka plan sekmesinde Durdur'a basar. |
 | `fetch_images` | `conversation_url` (zorunlu), `aspect_ratio`, `output_dir`, `wait_seconds` (45), `include_preview` (true) | Var olan bir sohbetin son cevabındaki görselleri, hiçbir şey göndermeden indirir. |
@@ -417,6 +446,8 @@ klasöründe kalır.
 - `send_as_is`: istemi, Backdraw'ın eklediği kısa talimat olmadan, olduğu gibi gönderir (örneğin
   ChatGPT'nin sorduğu bir soruyu yanıtlamak için).
 - `output_dir`: sonucu bu klasöre de kopyalar; görsel klasöründe her zaman bir kopya kalır.
+- `filename_hint`: konuyu anlatan birkaç kelime; dosya adı istemin başı yerine bundan üretilir
+  (istem referans görsellerle ilgili talimatlarla başlıyorsa işe yarar).
 - `wait_seconds`: 0 ile 900 arası.
 
 Bir sonuç şunları söyleyebilir:
@@ -446,7 +477,8 @@ beklemeyi durdurur; iş, `cancel_job` onu durdurana kadar sürer.
 
 Kurulum ve sorun giderme için bunları Backdraw klasöründe çalıştır (`%USERPROFILE%\.backdraw` ya da
 kurulumu yaptığın klasör). Her yerden `npx backdraw <komut>` aynı komutları çalıştırır, örneğin
-`npx backdraw status`. `npm run status`, `node cli.js status` ile aynıdır.
+`npx backdraw status`. `npm run status`, `node cli.js status` ile aynıdır. `backdraw mcp`, MCP
+sunucusunun kendisini stdio üzerinde çalıştırır; MCP istemcilerinin başlattığı budur.
 
 ```
 node cli.js status                          Köprünün durumu (Chrome bağlı mı, seçili proje, kuyruk)

package/bin/backdraw.js

@@ -16,6 +16,7 @@ Usage: backdraw <command>
                                   clients: claude-code, claude-desktop, codex, cursor, vscode, gemini
   doctor [--fix]                  Check the install; --fix repairs registrations
   uninstall [--dry-run]           Remove the registrations (your images and settings stay)
+  mcp                             Run the MCP server on stdio (what MCP clients start; needs setup once)
   --version                       Print the version
 ${cli.USAGE.split('\n').slice(1).join('\n')}`;
 
@@ -33,6 +34,9 @@ async function main(argv) {
   const [cmd, ...rest] = argv;
   const { flags } = parseArgs(rest, { booleans: ['yes', 'fix', 'dry-run', 'no-wait'] });
   switch (cmd) {
+    // stdout belongs to the JSON-RPC stream from here on; the server reads stdin until it closes.
+    case 'mcp':
+      return require('../mcp-server');
     case '-v': case '--version': case 'version':
       return console.log(VERSION);
     case 'setup':

package/extension/manifest.json

@@ -2,7 +2,7 @@
   "manifest_version": 3,
   "name": "Backdraw",
   "short_name": "Backdraw",
-  "version": "2.1.0",
+  "version": "2.1.2",
   "description": "__MSG_extDescription__",
   "key": "MIIBIjANBgkqhkiG9w0BAQEFAAOCAQ8AMIIBCgKCAQEAtu/jyAJ3e3hnNrNGCxfzvHLOp80L9c3j71jmQV2ymY1DbPfZNJUdrszM3dnk6SFlm/X42F/e1Pf9RoQqMKujWr7KRTtH257vBMmLbk8Eo0LytpGP0F8WOQIcp5TxLZgvqSlGTjdbB2QFJt+uisJFE8Y3nWnZpsmejes0m9NObpqBc7Q1tzZrDAnh6gzddyF8fkYU/gS2Qtu5Vd2YbRiVN18tUUh2Bb+QuXlTlzOE1JQ49A9qRTmXejtcn943ZaQ0aucooNQU4hZAFVUT+25UhzuGMZDFzZALqL19hP0EP4Erm1yitjICvR5yK/7nYObGa3LynEeH7dDJ+qoJaf4TEQIDAQAB",
   "default_locale": "en",

package/host.js

@@ -199,6 +199,8 @@ function createJob(p) {
     // outputDir: legacy clients, the host writes a copy there. copyTo: the MCP server copies itself
     // (it runs in the caller's context, so the copy lands where the caller can see it).
     outputDir: p.outputDir || null, copyTo: p.copyTo || null, copies: [],
+    // Names the saved files instead of the prompt (whose opening may be instructions, not the subject).
+    filenameHint: String(p.filenameHint || '').trim().slice(0, 200) || null,
     references: refs.map(r => ({ name: r.name, bytes: Math.round(r.data.length * 0.75) })), autoReattach: !!p.autoReattach,
     timeoutSec: clamp(Number(p.timeoutSec) || 600, 60, 1800),
     createdAt: now(), startedAt: null, finishedAt: null,
@@ -323,9 +325,9 @@ function uniquePath(dir, base, ext) {
 function saveImages(job, images, title) {
   fs.mkdirSync(config.outputDir, { recursive: true });
   fs.mkdirSync(shared.PREVIEW_DIR, { recursive: true });
-  // Named after the prompt; when only collecting (no prompt) ChatGPT's chat title is used instead.
+  // Named after filename_hint, else the prompt; when only collecting (no prompt) ChatGPT's chat title.
   const usefulTitle = title && !/^(chatgpt\b|generate an image|update the image)/i.test(title.trim()) ? title : '';
-  const base = `${stamp()}_${slugify(job.prompt || usefulTitle)}`;
+  const base = `${stamp()}_${slugify(job.filenameHint || job.prompt || usefulTitle)}`;
   return images.map((img, i) => {
     const buf = Buffer.from(img.data || '', 'base64');
     if (buf.length < 100) throw new Error('empty image data');

package/lib/results.js

@@ -15,6 +15,9 @@ const checkAgain = id => `Call job_status with job_id="${id}" and wait_seconds=$
 // A prompt that asks to change an existing image rather than draw a new one.
 const EDIT_INTENT = /\b(edit(s|ed|ing)?|modif(y|ies|ied|ying)|retouch(es|ed|ing)?|inpaint(s|ed|ing)?|change only|only change|change [a-z ]{1,30} only|keep (everything|it|the rest)|pixel-identical|same (pose|image|drawing|character))\b|düzenle|değiştir|aynı (kalsın|bırak|tut)|koru(yarak)?\b/i;
 
+const SHOW_IMAGE = 'Show the image to the user in the chat: if you can display or share a local file with the user, do that with the Saved path above ' +
+  '(any preview attached here is a small copy for you to check); otherwise give that path clearly.';
+
 const neutral = (pct, against) => `Similarity to ${against}: ${pct}% (a local edit keeps ~95-100%, a redraw is usually under 60%).`;
 
 function editCheckLines(job) {
@@ -156,6 +159,8 @@ function resultLines(job) {
   for (const im of job.images || []) lines.push(...imageLines(job, im));
   for (const c of job.copies || []) lines.push(`Copied to: ${c}`);
   if (job.copyError) lines.push(`Could not copy into output_dir: ${job.copyError}`);
+  // Clients differ: some show the attached preview only as a small icon, so ask for the file itself.
+  if (job.status === 'succeeded' && (job.images || []).length) lines.push(SHOW_IMAGE);
   lines.push(...editCheckLines(job));
   if (job.assistantText) {
     lines.push(job.status === 'succeeded' ? `ChatGPT's text: ${job.assistantText.slice(0, 1500)}` : `ChatGPT's reply:\n${job.assistantText.slice(0, 3000)}`);
@@ -167,4 +172,4 @@ function resultLines(job) {
   return lines;
 }
 
-module.exports = { FINAL, DEFAULT_WAIT, EDIT_INTENT, waitSeconds, editCheckLines, imageLines, referenceLines, resultLines };
+module.exports = { FINAL, DEFAULT_WAIT, EDIT_INTENT, SHOW_IMAGE, waitSeconds, editCheckLines, imageLines, referenceLines, resultLines };

package/mcp-server.js

@@ -19,12 +19,14 @@ const INSTRUCTIONS = `Generates images with the user's own ChatGPT account (alre
 - To EDIT an image or keep a character consistent, pass the image file(s) in reference_images (e.g. a file saved by an earlier call). A fresh chat with reference_images is the most reliable way to edit; long revision chains via conversation_url make ChatGPT lose track of earlier images.
 - For a specific shape (banner, avatar, phone wallpaper) pass aspect_ratio instead of describing the size in the prompt; for stickers/logos/icons that need a see-through background pass transparent_background=true and read the transparency line in the result.
 - When the user mentions one of their Backdraw collections (reference images they saved from the web in Chrome), call list_collections if needed and pass reference_collection to generate_image.
+- The saved file is named after the start of the prompt. When the prompt opens with something other than the subject (e.g. instructions about the reference images), pass filename_hint with a few words naming what the user asked for.
 - Read every result line: NEEDS REFERENCE IMAGE means ChatGPT could not reach the image to edit; CHATGPT REPLIED WITHOUT AN IMAGE shows ChatGPT's own text. Never resend the identical prompt in a loop; follow the next step given in the result.
 - An EDIT CHECK warning means ChatGPT drew a new picture instead of editing; tell the user and retry with reference_images in a fresh chat if an exact edit is needed.
 - The work happens in a background Chrome tab; never tell the user to switch tabs unless a result says NEEDS USER ACTION.
 - NEEDS USER ACTION means ChatGPT shows a login or human-verification (CAPTCHA) page: ask the user to open the ChatGPT tab in Chrome and finish it themselves. Never try to get around it. The job continues automatically afterwards (check with job_status).
 - If no project is selected, call list_projects, then select_project with the user's images project.
 - If a job timed out but has a conversation link, fetch_images can still download the result later; pass the job's aspect_ratio again so it is cropped the same way.
+- After SUCCESS, show the saved image to the user in the chat: if you can display or share a local file with the user, do that with the saved path (the preview in the result is a small copy for you to check); otherwise give the saved path clearly.
 - Report the saved file path(s) and the ChatGPT conversation link to the user.`;
 
 // Shared by the tools that wait for a job.
@@ -78,6 +80,11 @@ const TOOLS = [
           type: 'string',
           description: "Optional folder to also copy the image(s) into, e.g. the current project's assets folder (absolute path preferred). The central images folder always keeps a copy.",
         },
+        filename_hint: {
+          type: 'string',
+          description: "Optional few words naming the subject (e.g. 'red fox mascot'), used for the saved file name instead of the start of the prompt. " +
+            'Pass it when the prompt opens with something else, such as instructions about the reference images.',
+        },
         wait_seconds: WAIT_SECONDS,
         include_preview: { type: 'boolean', default: true, description: 'Attach a downscaled preview of the result so you can check it.' },
         send_as_is: { type: 'boolean', default: false, description: 'Send the prompt verbatim, without the short instruction the bridge adds (useful to answer a question ChatGPT asked).' },
@@ -302,6 +309,7 @@ async function generateImage(args, call) {
     prompt, references, autoReattach: !!args.auto_reattach, sendAsIs: !!args.send_as_is,
     aspect: args.aspect_ratio, transparent: !!args.transparent_background,
     copyTo: args.output_dir ? path.resolve(String(args.output_dir)) : undefined,
+    filenameHint: args.filename_hint ? String(args.filename_hint) : undefined,
     startUrl: args.conversation_url ? String(args.conversation_url) : undefined,
     referenceCollection: args.reference_collection ? String(args.reference_collection) : undefined,
     // A single file name as a string works too, like reference_images.

package/package.json

@@ -1,7 +1,8 @@
 {
   "name": "backdraw",
-  "version": "2.1.0",
+  "version": "2.1.2",
   "description": "Local MCP server, native messaging host and Chrome extension that let Claude generate images with your own signed-in ChatGPT account in a background tab.",
+  "mcpName": "io.github.devtalh/backdraw",
   "license": "MIT",
   "author": "devtalh",
   "homepage": "https://github.com/devtalh/backdraw#readme",
@@ -49,8 +50,5 @@
     "status": "node cli.js status",
     "install-bridge": "node install.js",
     "test": "node --test"
-  },
-  "os": [
-    "win32"
-  ]
+  }
 }

package/setup.js

@@ -3,7 +3,7 @@
 // own copy is temporary), registers the native host for the installed browsers, adds Backdraw to
 // the MCP clients the user picks, then walks through loading the extension and waits for it.
 //   --yes                 no questions: every client that was found
-//   --clients=a,b         exactly these (ids: see CLIENT_IDS)
+//   --clients=a,b         exactly these (ids: see CLIENT_IDS); --clients=none: none, the browser side only
 //   --browser=edge        the browser to load the extension in (default: the first one found)
 //   --no-wait             don't wait for the extension to connect
 const fs = require('fs');
@@ -31,6 +31,7 @@ function detectClients(c) {
 function parseClientList(value) {
   if (value === true || value === '') throw new Error(`--clients needs a list, e.g. --clients=claude-code,cursor (one of: ${CLIENT_IDS.join(', ')}).`);
   const ids = String(value).split(',').map(s => s.trim().toLowerCase()).filter(Boolean);
+  if (ids.length === 1 && ids[0] === 'none') return []; // the browser side only (e.g. the Claude Code plugin)
   const unknown = ids.filter(id => !CLIENT_IDS.includes(id));
   if (unknown.length) throw new Error(`Unknown client ${unknown.join(', ')}; use: ${CLIENT_IDS.join(', ')}.`);
   return [...new Set(ids)];