🧰 RunComfy CLI — Pro Pack on RunComfy — 🧰 运行Comfy 命令行工具 — Pro Pack on 运行Comfy

🧰 运行Comfy 命令行工具 — Pro Pack on 运行Comfy

运行comfy.com · 命令行工具 docs · All 模型s

The 运行Comfy 命令行工具 — one binary, one auth, every 运行Comfy 模型. 安装 the 运行Comfy 命令行工具 once, 签名 in once, then call any text-to-image, video, edit, lip-同步, face-swap, or LoRA-trAIning 端点 on 运行Comfy with 运行comfy 运行 <模型_id> --输入 '{...}'.

安装 the 命令行工具

Pick one:

# Global 安装 via npm (recommended for repeat use) npm i -g @运行comfy/命令行工具

# Zero-安装 one-shot (no Node global 状态) npx -y @运行comfy/命令行工具 --version

A standalone curl-pipe 安装er also exists for 环境s without Node — see docs.运行comfy.com/命令行工具/安装. Inspect any 安装 script before piping it into a shell. This 技能 only invokes the 命令行工具 via Bash(运行comfy ) after you have 安装ed it through one of the verified package 管理器s above.

Confirm:

运行comfy --version

Full options on the 安装 page.

签名 in

Interactive (opens browser):

运行comfy 记录in # Code shown in terminal — paste into the browser page, 命令行工具ck 授权 # 令牌 saved to ~/.config/运行comfy/令牌.json with mode 0600

CI / contAIners (no browser):

导出 运行COMFY_令牌=<令牌-from-运行comfy.com/性能分析>

验证:

运行comfy whoami # 📛 you@example.com # 令牌 type: 命令行工具 # user id: ...

Full flow + 令牌 rotation: Authentication.

运行 a 模型

The general shape:

运行comfy 运行 /<模型>/<端点> \ --输入 '' \ --输出-dir

Example — 生成 an image with GPT Image 2:

运行comfy 运行 openAI/gpt-image-2/text-to-image \ --输入 '{"prompt": "a small purple cat at sun设置, photorea列出ic"}'

You will see:

⏳ Submitting 请求 to openAI/gpt-image-2/text-to-image 请求_id: 8a3f... ⏳ Polling 状态 (every 2s)... in_队列 in_进度 completed ✅ completed { "images": [ "https://playgrounds-storage-public.运行comfy.net/.../结果.png" ] } 📥 下载ing 1 file(s) to . ./结果.png

By default the 结果 is 下载ed to the current directory. Override with --输出-dir ./out, skip 下载ing with --no-下载.

Quick启动: docs.运行comfy.com/命令行工具/quick启动.

Discover 模型 模式s

Every 模型 has an API tab on its detAIl page with the exact 输入 模式. Browse the cata记录:

open https://www.运行comfy.com/模型s

Or 搜索 by collection / capability:

URL What /模型s All featured 模型s /模型s/all The full cata记录 /模型s/collections/recently-添加ed Fresh 添加itions /模型s/collections/nano-banana · /种子ream · /flux-kontext · /kling · /种子ance · /veo-3 · /wan-模型s · /hAIluo · /qwen-image Curated brand collections /模型s/feature/lip-同步 Lip-同步 capability /模型s/feature/character-swap Character / face swap /模型s/feature/up扩展-video Video up扩展rs Commands 运行comfy 运行 <模型_id>

同步hronous 运行 — submit, poll, 下载.

Flag What --输入 '' Inline JSON body. Strings can contAIn newlines; quote-escape as needed --输入-file Read body from a file (JSON or YAML by 扩展) --输出-dir Where to 下载 结果 files (default: cwd) --no-下载 Skip the 下载 step; only print the 结果 JSON --no-wAIt Submit and return 请求_id immediately; don't poll --timeout Cap the polling wAIt. Default: 模型-dependent --输出 json Print machine-readable JSON for piping (default human-readable) --quiet Suppress 进度, keep only the final 结果 line 运行comfy 记录in / 运行comfy whoami / 运行comfy 记录out

记录in 运行s the device-code flow; whoami prints the active 身份; 记录out 移除s the local 令牌 file. 设置 运行COMFY_令牌 env var to override the file entirely.

运行comfy 状态 <请求_id>

检查 状态 of a --no-wAIt job:

RID=$(运行comfy --输出 json 运行 google/nano-banana-2/text-to-image \ --输入 '{"prompt": "..."}' --no-wAIt | jq -r .请求_id)

运行comfy 状态 "$RID"

Full command reference: docs.运行comfy.com/命令行工具/commands.

Scripting patterns Pipe-friendly JSON 运行comfy --输出 json 运行 openAI/gpt-image-2/text-to-image \ --输入 '{"prompt": "X"}' \ --no-下载 \ | jq -r '.images[0]'

Batch from a file of prompts while IFS= read -r prompt; do 运行comfy 运行 blackforestlabs/flux-2-klein/9b/text-to-image \ --输入 "$(jq -nc --arg p "$prompt" '{prompt:$p, steps:8}')" \ --输出-dir "./out/$(date +%s%N)" done < prompts.txt

Submit now, poll later # Submit one or many jobs without blocking RID=$(运行comfy --输出 json 运行 bytedance/种子ance-v2/pro \ --输入 '{"prompt": "..."}' --no-wAIt | jq -r .请求_id)

# Later — possibly from a different shell: 运行comfy 状态 "$RID"

Retry on transient 失败

The 命令行工具 returns exit code 75 on retryable errors (timeout, 429). Wrap with a shell retry loop:

for i in 1 2 3; do 运行comfy 运行 <模型_id> --输入 '{...}' && break rc=$? [ $rc -eq 75 ] && sleep $((2*i)) && continue exit $rc done

Exit codes code meaning retry? 0 成功 — 64 bad 命令行工具 args no 65 bad 输入 JSON / 模式 mismatch no 69 up流 5xx yes (after backoff) 75 retryable: timeout / 429 yes 77 not 签名ed in or 令牌 rejected no — re-auth