Open Source•Apache-2.0 License

badgr-cli

Run GPU jobs and serve OpenAI-compatible model endpoints from one command — with spend caps, live logs, automatic teardown, and receipts.

npm install -g badgr-cliGitHub →

How it works

badgr wraps GPU provisioning, execution, and teardown behind two commands — run for one-off jobs, serve for persistent endpoints — with cost safety built in from the start.

1

Run or serve

badgr run . --cmd "python train.py" --max-cost 5 for one-off jobs; badgr serve <model> --max-cost 10 for a persistent OpenAI-compatible endpoint.

2

Live logs + spend cap

Every job streams logs and auto-stops once --max-cost is reached. badgr down <id> stops billing immediately at any time.

3

Receipts

badgr receipts returns cost, route, retry, and teardown history for every run and serve action.

1

Install

Terminal
npm install -g badgr-cli

Requires Node.js 18+ and a Badgr account (sign up at aibadgr.com).

2

Usage

Authenticate once, then run a project folder on a GPU with a hard cost cap:

Terminal
badgr login
badgr run . --cmd "python train.py" --max-cost 5

Commands

badgr loginSave your API key to ~/.badgr/config.json
badgr connect <provider>Store a provider credential (anthropic, openai) for badgr launch
badgr doctorDiagnose a GPU workload failure — read-only, no login needed
badgr run <command>Run a one-off GPU job (any container command)
badgr launch cline|claude|codex|playwright "<task>"Run a coding/testing agent on a CPU VM — image + command auto-selected
badgr task "<desc>" -- <command>Label + launch — thin wrapper over badgr launch . -- <command>
badgr serve <model>Start a persistent OpenAI-compatible endpoint
badgr statusShow what's running and what's billing
badgr logs <id>Fetch log output from a deployment
badgr pull <id>Pull a code-editing agent's patch as a local git diff/branch
badgr artifacts <id>Download non-patch outputs (test reports, screenshots, traces)
badgr down <id>Terminate a deployment — stops billing immediately
badgr receipts [n]Cost, route, and retry receipts (default 10)
badgr batch run <workload.yml>Run a generic containerized batch job with artifact capture
badgr sbatch <job.slurm>Run an existing Slurm batch script (cpus/mem/gres/array/time translated)
badgr testRun an end-to-end test (provision → run → teardown)

What's included

✓badgr run for one-off GPU jobs from a local folder, a public GitHub repo, or a custom Docker image
✓badgr launch cline/claude/codex/playwright for coding and testing agents on a CPU VM — no --max-cost required, image and command auto-selected
✓badgr connect anthropic/openai to link your own account for badgr launch claude/codex (Cline uses Badgr-provided model access; Playwright needs no account) — also prompted inline if missing
✓badgr pull and badgr artifacts to retrieve a coding agent's patch or any other captured output (test reports, screenshots, traces)
✓badgr serve for persistent OpenAI-compatible vLLM endpoints, including blessed model aliases
✓badgr batch run for generic containerized batch jobs, with --fan-out to run one deployment per input file in parallel
✓badgr sbatch to run an existing Slurm script — cpus/mem/gres/array/time translated, array jobs fan out automatically
✓badgr comfyui batch for productized image generation through blessed ComfyUI workflows — no setup needed
✓badgr doctor for read-only diagnosis of GPU, driver, CUDA, and PyTorch failures — no login required
✓badgr train and badgr train lora for Axolotl-based fine-tuning and LoRA adapter training
✓badgr transcribe and badgr embed for Whisper transcription and vLLM text embeddings
✓Required --max-cost spend caps on run/comfyui/train jobs, plus --dry-run previews before provisioning
✓badgr receipts with cost, route, retry, and teardown history for every job
✓Saved badgr workload and badgr workspace tracking for repeatable jobs and cost/context grouping
View on GitHub →

Ready to try badgr-cli?

npm install -g badgr-cli — no account needed to get started.

Install from npmGet cloud API key