runcomfy CLI

Command-line tool for RunComfy: run hosted AI media models, manage Serverless API (ComfyUI) deployments, and train LoRAs — the same surface as the RunComfy MCP server, from a terminal or a script.

Install

# Easiest — via npx (no install)
npx -y @runcomfy/cli login

# Or install globally via npm
npm i -g @runcomfy/cli

# Or curl install (Rust binary directly)
curl -fsSL https://runcomfy.com/install.sh | sh

Quick start

# Login (opens browser for OAuth); or export RUNCOMFY_TOKEN in CI
runcomfy login
runcomfy whoami
runcomfy balance

# Find a hosted model, read its input schema, run it
runcomfy models list --search kontext
runcomfy models get blackforestlabs/flux-1-kontext/pro/edit
runcomfy run blackforestlabs/flux-1-kontext/pro/edit \
  --input '{"prompt": "she is now holding an umbrella", "image_url": "https://..."}'

# Submit without waiting; fetch the result (and files) later
runcomfy --output json run openai/gpt-image-2/text-to-image \
  --input '{"prompt": "ukiyo-e mountain"}' --no-wait | jq -r .request_id
runcomfy status <request_id>
runcomfy result <request_id> --output-dir ./out

# Serverless API (ComfyUI) deployments
runcomfy deployments list
runcomfy deployments get <deployment_id> --include-payload   # node IDs for overrides
runcomfy deployments run <deployment_id> \
  --overrides '{"6": {"inputs": {"text": "a futuristic city"}}}'

# LoRA training: dataset -> job -> checkpoints
runcomfy datasets create --name my-dataset
runcomfy datasets upload <dataset_id> ./my-dataset/ --wait
runcomfy train submit --config ./config.yaml
runcomfy train status <job_id>
runcomfy train result <job_id> --download --output-dir ./lora

Every command accepts --output json for machine-readable output, -q, and -v. Exit codes follow sysexits(3); see runcomfy --help.

Commands

Group	Commands	Backs
Auth	login, logout, whoami	www.runcomfy.com/api/cli-auth, /api/auth/me
Account	balance	GET /prod/v2/balance
Model catalog	models list, models categories, models get	Model API GET /v1/models…
Model API requests	run, status, result, cancel (aliases: requests get/result/cancel)	Model API POST /v1/models/{id}, GET/POST /v1/requests/{id}/…
Serverless deployments	deployments list/get/create/update/delete	GET/POST/PATCH/DELETE /prod/v2/deployments…
Serverless inference	deployments run/status/result/cancel/proxy	POST …/inference, …/requests/{id}/…, …/instances/{id}/proxy/…
Trainer datasets	datasets create/list/status/delete/upload	…/trainers/datasets… (multipart ≤150 MB, signed URLs above)
Trainer jobs	train submit/status/result/cancel/resume/edit	…/trainers/ai-toolkit/jobs…

Waiting commands (run, deployments run) cancel the remote request on Ctrl-C so an abandoned run stops billing. train submit --wait and datasets upload --wait only stop watching on Ctrl-C — the job keeps running. Destructive commands (deployments delete, datasets delete) prompt, or need --yes when not attached to a terminal.

Generated files are downloaded only from RunComfy CDN hosts (*.runcomfy.net, *.runcomfy.com); other URLs are listed but not fetched.

Build from source

cargo build --release
./target/release/runcomfy --help

Architecture

cli/
├── Cargo.toml
├── src/
│   ├── main.rs        Entry point + tokio runtime
│   ├── cli.rs         clap subcommand definitions
│   ├── config.rs      Local config + token persistence (~/.config/runcomfy/)
│   ├── api.rs         HTTP client + base URLs for the RunComfy APIs
│   ├── poll.rs        Wait-until-terminal loop (Ctrl-C cancel / detach)
│   ├── download.rs    Trusted-host asset downloads
│   ├── input.rs       JSON / file / stdin argument parsing
│   ├── output.rs      --output pretty|json, progress, tables
│   ├── error.rs       Error types
│   ├── exit.rs        sysexits-style exit codes
│   └── cmd/           Subcommand implementations
│       ├── login.rs / whoami.rs / balance.rs
│       ├── models.rs
│       ├── run.rs / status.rs / result.rs / cancel.rs
│       ├── deployments.rs
│       ├── datasets.rs
│       └── train.rs

Endpoints

• Model API: https://model-api.runcomfy.net/v1/... (RUNCOMFY_MODEL_API_BASE)
• Serverless API: https://api.runcomfy.net/prod/v2/... (RUNCOMFY_SERVERLESS_API_BASE)
• Trainer API: https://trainer-api.runcomfy.net/prod/v1/... (RUNCOMFY_TRAINER_API_BASE)
• Web auth: https://www.runcomfy.com/api/cli-auth/... (RUNCOMFY_WEB_BASE)

Docs: https://docs.runcomfy.com/cli/introduction