E
Ekko237/imagegen-swarm
Use every AI image quota you already pay for, all at once: a Claude Code / Codex skill that runs Codex, Grok,…
在 GitHub 查看 ↗链接已于 2026-11-12 核查 · HTTP 200 · 打开的是厂商自己的官网
安装与配置
客户端配置（取自该仓库的安装命令）
{
"mcpServers": {
"ekko237-imagegen-swarm": {
"command": "npx",
"args": ["skills"]
}
}
}解析自我们于 2026-11-11 抓取的该仓库 README；我们未执行过。
command/args 来自仓库写明的安装命令；不确定时请以该项目 README 为准。
VS Code（.vscode/mcp.json —— 键名是 servers）
{
"servers": {
"ekko237-imagegen-swarm": {
"command": "npx",
"args": ["skills"]
}
}
}VS Code 顶层键是 <code>servers</code>，其余与上面那份相同。
Claude Code（命令行）
claude mcp add --transport stdio ekko237-imagegen-swarm -- npx skills
Claude Code 不收 JSON 文件，用这条命令直接加。
配置项的键名以你的客户端为准：客户端与 CLI
- ★ 数取值
- 2 ★ · Ekko237/imagegen-swarm · 取自 GitHub 搜索 API · （2026-11-13 20:00:50 核对）
- 来源明细
- 从该仓库的 GitHub 页面直接读取 · 条目类型：开源项目
- 收录与活跃
- 收录于 2026-11-12 · 最近见到 2026-11-14 · 共见到 3 次 · 被 1 个来源列出
- 评分构成
- 综合分 103 · 2 ★ · 1 个来源 · repo
- 安装
npx skills
常见问题（答案都来自我们的记录）
怎么安装？
该仓库给出的安装命令是
npx skills。我们的记录里它是仓库自己提供的；不保证在你的客户端上直接可用。它适合谁 / 用来做什么？
我们把它归在这些标签下: Agent 技能。标签来自条目自身，不是我们强加的分类。
谁维护它、还在更新吗？
仓库
Ekko237/imagegen-swarm；我们收录于 2026-11-12；最近见到 2026-11-14。这里没有"官方认证"或"质量评级"——我们只记录来源与时间。# imagegen-swarm
以下摘录自该仓库的 README（片段） · 在 GitHub 读完整 README →
中文说明
A Claude Code and Codex skill that runs batch image and video generation across six engines in parallel. No image API key needed for most engines—use the Codex, Jimeng, Grok, Gemini, and Muse accounts you already have.
Codex, Jimeng, Grok, Gemini, Muse: use your existing account logins. (Jimeng also generates videos.)
Lovart: uses its official AK/SK, stored on your machine.
Give imagegen-swarm a list of jobs and it feeds all six engines at once. Free lanes pull the next image from a shared queue, finishing batches in minutes. When one engine hits its quota limit, the remaining engines pick up its work.
## How the work is shared
Each engine opens multiple lanes. Every free lane grabs the next job in the queue. If an engine runs out of quota, the others take over. In the 50-image test run above, Gemini and Lovart hit their limits partway through, and the remaining engines still delivered all 50 images.
It is one test run, so your numbers will differ: your speed will depend on your account quotas. In this run, Lovart and Muse were on free plans, and Gemini and Lovart ran out midway. With more quota, all engines stay active to the end and a batch usually finishes faster.
## Why
Image tools give you weekly or monthly quotas that mostly sit unused. Running one engine at a time is slow and stops at its limit. imagegen-swarm puts all your accounts on a single queue:
Parallel: every engine runs multiple lanes at once; any free lane grabs the next compatible job.
Quota-aware: if an engine reports exhausted quota, a missing login, or a missing install, it shuts down and returns its job to the queue. Plain rate limits trigger a backoff and only shut the engine down after three in a row.
Correct: checks every file to confirm it is new from this run, re-runs wrong aspect ratios on another engine, and output is always real PNG.
Fits your machine: lane counts scale with RAM, from 12 lanes on a 4 GB laptop to 34 on 16 GB or more. Want to keep memory free for other apps? Cap memory usage with IMAGE_GEN_MEMORY_GB=4, or set exact lane counts with --lanes gemini=2,codex=2.
Resumable: rerun the same job list and already-finished images are skipped.
## Engines
| Engine | Can do | Runs through |
| codex | images, reference images | Codex CLI's built-in image_gen (your Codex / ChatGPT login) |
| jimeng | images, edits (up to 10 references), text- and image-to-video | the official Dreamina CLI (your Douyin login) |
| grok | images, edits, reference-to-video | Grok Build CLI's Imagine tools |
| gemini | images, edits (up to 3 references) | Antigravity CLI's generate_image |
| muse | images, reference images | Muse (muse.ai) driven by a headless browser |
| lovart | images, reference images | the official Lovart CLI (your AK/SK) |
You only need one engine logged in. Every engine you add makes batches faster.
## Install
bash
npx skills add https://github.com/Ekko237/imagegen-swarm -a claude-code -a codex
That installs the skill with the npx skills CLI and works across Claude Code, Codex and other agents.
Claude Code plugin
Claude Code:
bash
claude plugin install imagegen-swarm --marketplace Ekko237/imagegen-swarm
中文说明
A Claude Code and Codex skill that runs batch image and video generation across six engines in parallel. No image API key needed for most engines—use the Codex, Jimeng, Grok, Gemini, and Muse accounts you already have.
Codex, Jimeng, Grok, Gemini, Muse: use your existing account logins. (Jimeng also generates videos.)
Lovart: uses its official AK/SK, stored on your machine.
Give imagegen-swarm a list of jobs and it feeds all six engines at once. Free lanes pull the next image from a shared queue, finishing batches in minutes. When one engine hits its quota limit, the remaining engines pick up its work.
## How the work is shared
Each engine opens multiple lanes. Every free lane grabs the next job in the queue. If an engine runs out of quota, the others take over. In the 50-image test run above, Gemini and Lovart hit their limits partway through, and the remaining engines still delivered all 50 images.
It is one test run, so your numbers will differ: your speed will depend on your account quotas. In this run, Lovart and Muse were on free plans, and Gemini and Lovart ran out midway. With more quota, all engines stay active to the end and a batch usually finishes faster.
## Why
Image tools give you weekly or monthly quotas that mostly sit unused. Running one engine at a time is slow and stops at its limit. imagegen-swarm puts all your accounts on a single queue:
Parallel: every engine runs multiple lanes at once; any free lane grabs the next compatible job.
Quota-aware: if an engine reports exhausted quota, a missing login, or a missing install, it shuts down and returns its job to the queue. Plain rate limits trigger a backoff and only shut the engine down after three in a row.
Correct: checks every file to confirm it is new from this run, re-runs wrong aspect ratios on another engine, and output is always real PNG.
Fits your machine: lane counts scale with RAM, from 12 lanes on a 4 GB laptop to 34 on 16 GB or more. Want to keep memory free for other apps? Cap memory usage with IMAGE_GEN_MEMORY_GB=4, or set exact lane counts with --lanes gemini=2,codex=2.
Resumable: rerun the same job list and already-finished images are skipped.
## Engines
| Engine | Can do | Runs through |
| codex | images, reference images | Codex CLI's built-in image_gen (your Codex / ChatGPT login) |
| jimeng | images, edits (up to 10 references), text- and image-to-video | the official Dreamina CLI (your Douyin login) |
| grok | images, edits, reference-to-video | Grok Build CLI's Imagine tools |
| gemini | images, edits (up to 3 references) | Antigravity CLI's generate_image |
| muse | images, reference images | Muse (muse.ai) driven by a headless browser |
| lovart | images, reference images | the official Lovart CLI (your AK/SK) |
You only need one engine logged in. Every engine you add makes batches faster.
## Install
bash
npx skills add https://github.com/Ekko237/imagegen-swarm -a claude-code -a codex
That installs the skill with the npx skills CLI and works across Claude Code, Codex and other agents.
Claude Code plugin
Claude Code:
bash
claude plugin install imagegen-swarm --marketplace Ekko237/imagegen-swarm
收录徽章（可选）
想把收录标记放进 README 就复制这行，可选，不影响收录。
[](https://anclum.com/zh/s/ekko237-imagegen-swarm/)
- 来源
- github · https://github.com/Ekko237/imagegen-swarm
- 按领域浏览
- Agent 技能
- Traction
- 2 ★
- 收录于
- 今天
同类推荐
所属领域与标签
每个标签都能点进去看同类的全部条目。数字是当前库里的真实条数。