By vast-ai
Manage Vast.ai GPU instances, volumes, serverless endpoints, and billing from Claude Code. Natural-language driven via the bundled vastai skill, with five slash commands for hot-path operations.
Snapshot of Vast.ai spend — account balance, active $/hr, recent invoices
Launch a Vast.ai instance from an offer ID with sane defaults (PyTorch image, --ssh --direct --cancel-unavail)
Search Vast.ai GPU offers by filter — returns the cheapest matching rentable offers
First-time Vast.ai setup — store API key, register SSH key, verify auth
Snapshot of Vast.ai instance status — single instance or all running
Vast.ai CLI for GPU hosts/providers — list and unlist machines on the marketplace, set pricing (min-bid, default GPU price), configure default jobs, schedule maintenance windows, run self-tests, view earnings, monitor marketplace metrics (gpu, gpu-trends, gpu-locations), manage network disks and clusters, defrag machines, clean up expired storage. Use this for any prompt about hosting on Vast, listing a machine, machine pricing, maintenance windows, host earnings, marketplace metrics, or any vastai command for GPU providers.
Vast.ai CLI for renters — search and launch GPU instances, SSH into them, copy files, run commands, manage volumes and serverless endpoints, manage user environment variables (HF_TOKEN, OPENAI_API_KEY, model config), check billing and balance, register SSH keys, destroy instances. Use this for any prompt about Vast.ai, vastai, GPU rental, or instance lifecycle.
Runs pre-commands
Contains inline bash commands via ! syntax
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
A Claude Code plugin for Vast.ai. Rent, launch, monitor, and tear down GPU instances — plus volumes, serverless endpoints, and billing — without leaving your editor.
The plugin combines two layers:
vastai CLI, so anything the slash commands don't cover works through natural language:
vastai — renter operations: search and launch instances, SSH, copy, logs, exec, destroy, volumes, serverless, env vars, billing.vastai-host — GPU provider operations: list/unlist machines, pricing, maintenance windows, self-tests, earnings, marketplace metrics. Auto-loads on host-intent prompts.| Command | What it does |
|---|---|
/vastai:setup [api-key] | Stores your API key, registers your SSH public key with Vast, and verifies the credential against vastai show user. Run this once before anything else — launching an instance before an SSH key is registered produces an unreachable host. |
/vastai:status [instance-id] | Snapshot of a single instance or all of yours. Flags terminal states (exited, unknown, offline) so polling loops actually terminate. |
/vastai:cost | Account balance, current burn rate ($/hr across active instances), and projected 24-hour spend. Useful as a pre-flight check before launching anything large. |
/vastai:search [filter] | Finds the cheapest rentable offers matching a filter, sorted by total $/hr. Recognises "spot" / "bid" in the filter and switches to interruptible bid-mode pricing automatically. |
/vastai:launch <offer-id> [image] [--disk N] [--label STR] | Launches an offer with sensible defaults: pytorch/pytorch:@vastai-automatic-tag, 20 GB disk, direct SSH, JSON output. Parses the returned new_contract ID so subsequent commands can reference it. |
Every vastai invocation includes --raw so responses come back as parseable JSON.
Anything outside the slash-command set works conversationally. The right skill auto-loads based on intent.
Renter prompts load vastai:
"Find a verified 4090 under $0.40/hr and launch it with pytorch."
"SSH into instance 12345 and tail
/var/log/nvidia-installer.log.""Set
HF_TOKENtohf_xxxxxon my running instance.""Destroy everything except instance 12345."
Host prompts load vastai-host:
"List my machine 98765 at $0.30/GPU/hr."
"Schedule maintenance on machine 98765 next Tuesday at 02:00 UTC for 4 hours."
"Show my host earnings for last month."
"What's the going rate for RTX 4090s in the US right now?"
To force a skill to load explicitly, mention it by name ("using the vastai skill, …" or "using the vastai-host skill, …").
pip install vastai
/plugin install vastai@claude-plugins-official
git clone https://github.com/vast-ai/vast-claude-plugin.git
claude --plugin-dir ./vast-claude-plugin
/vastai:setup
You'll be prompted for an API key from https://console.vast.ai/manage-keys/. The command then registers your SSH public key (~/.ssh/id_ed25519.pub by default, falling back to id_rsa.pub) and verifies the credential by querying your user record.
vast-claude-plugin/
├── .claude-plugin/plugin.json # marketplace manifest
├── commands/
│ ├── setup.md # /vastai:setup
│ ├── status.md # /vastai:status
│ ├── cost.md # /vastai:cost
│ ├── search.md # /vastai:search
│ └── launch.md # /vastai:launch
├── skills/
│ ├── vastai/SKILL.md # renter skill
│ └── vastai-host/SKILL.md # GPU provider / host skill
└── TESTING.md # acceptance test plan
MIT — see LICENSE.
npx claudepluginhub vast-ai/vast-claude-plugin --plugin vastaiComprehensive UI/UX design plugin for mobile (iOS, Android, React Native) and web applications with design systems, accessibility, and modern patterns
Standalone image generation plugin using Nano Banana MCP server. Generates and edits images, icons, diagrams, patterns, and visual assets via Gemini image models. No Gemini CLI dependency required.
Multi-model consensus engine integrating OpenAI Codex CLI, Gemini CLI, and Claude CLI for collaborative code review and problem-solving.
Write feature specs, plan roadmaps, and synthesize user research faster. Keep stakeholders updated and stay ahead of the competitive landscape.