WaveSpeed Agent Skills
Agent Skills for generating and editing AI media — image, video, audio, 3D — on the WaveSpeed platform. One general skill teaches the agent the image and video shortcuts and the find → inspect → run workflow, on the open-source wavespeed CLI or the WaveSpeed MCP tools; twenty-one per-model skills carry each model's inputs, pricing, and prompt tips. The bundled @wavespeed/mcp server exposes the same platform as MCP tools.
Every skill follows the Agent Skills SKILL.md format, and the repo root is an Agent Plugins 1.0 package (plugin.json + mcp.json + skills/), so it installs as-is into Kiro, Codex, Qwen Code, OpenHands and any other Agent Plugins consumer.
Skills
| Skill | What it covers |
|---|---|
wavespeed | The whole catalog: search models, read a model's schema, quote the price, run, upload local files with @path. Start here. |
wavespeed-seedream-5-pro | ByteDance Seedream 5 Pro text-to-image, editing and layer decomposition |
wavespeed-gpt-image-25 | OpenAI GPT Image 2.5 (Sunburst and Flare) text-to-image and editing |
wavespeed-nano-banana-21 | Google Nano Banana 2.1 text-to-image and editing |
wavespeed-seedance-25 | ByteDance Seedance 2.5 text, image and reference to video, video edit and extend |
wavespeed-kling-40 | Kuaishou Kling 4.0 text, image and reference to video, video edit and extend |
wavespeed-wan-30 | Alibaba Wan 3.0 and Wan 3.0 Prime video generation, edit and extend |
wavespeed-minimax-h3 | MiniMax H3 video generation, edit and extend, 768p by default |
wavespeed-minimax-speech-28 | MiniMax Speech 2.8 HD and Turbo text-to-speech |
wavespeed-seedream-45 | ByteDance Seedream 4.5 text-to-image and image editing |
wavespeed-nano-banana-pro | Google Nano Banana Pro text-to-image and editing, native 4K |
wavespeed-nano-banana-2 | Google Nano Banana 2 text-to-image and editing |
wavespeed-seedance-15-pro | ByteDance Seedance 1.5 Pro text-to-video and image-to-video |
wavespeed-veo-31-fast | Google Veo 3.1 Fast text-to-video, image-to-video, video extend |
wavespeed-wan-26 | Alibaba Wan 2.6 text-to-video and image-to-video |
wavespeed-wan-22-animate | Wan 2.2 Animate character animation and replacement |
wavespeed-infinitetalk-avatar | InfiniteTalk talking-head video from a portrait and audio |
wavespeed-minimax-speech-26 | MiniMax Speech 2.6 text-to-speech |
wavespeed-image-upscaler | Image upscaling to 2K / 4K / 8K |
wavespeed-ultimate-video-upscaler | Video upscaling to 720p / 1080p / 2K / 4K |
wavespeed-face-swapper | Image and video face swap, with consent and misuse guardrails |
wavespeed-watermark-remover | Watermark and overlay removal for media you own, with rights guardrails |
Install
Any agent that reads .agents/skills/ or .claude/skills/ (Claude Code, Codex, Cursor, OpenCode, Kimi Code, Antigravity, Copilot CLI, Amp, Droid, ...):
npx skills add WaveSpeedAI/agent-skills # all skills
npx skills add WaveSpeedAI/agent-skills --skill wavespeed
Kiro — Powers panel → Add Custom Power → Import power from GitHub → https://github.com/WaveSpeedAI/agent-skills.
Qwen Code / Codex / Kimi Code — the repo is an Agent Plugins package: qwen extensions install WaveSpeedAI/agent-skills, or the dedicated packages below.
OpenClaw — every skill is on ClawHub under @wavespeed: clawhub install wavespeed.
Dedicated packages (same skills, native packaging):
- Claude Code: WaveSpeedAI/claude-plugins
- Codex: WaveSpeedAI/codex-plugin-wavespeed-cli
- Kimi Code CLI: WaveSpeedAI/wavespeed-kimi-plugin
- Gemini CLI: WaveSpeedAI/wavespeed-gemini-extension
- DeepSeek Harness: WaveSpeedAI/wavespeed-dsh-skill
Setup
npm install -g @wavespeed/cli
wavespeed login # opens https://wavespeed.ai/accesskey and stores the key
WAVESPEED_API_KEY in the environment also works, and the MCP server shares the CLI's stored login. The skills tell the agent never to ask a user to paste a key into the chat.
MCP
mcp.json starts @wavespeed/mcp over stdio with npx. Tools: list_models, get_model_schema, get_price, upload_file, run_model, generate_image, edit_image, generate_video, edit_video, extend_video, get_prediction, get_balance. Each skill's CLI examples map one-to-one onto run_model with the same model id and inputs.
Recommended models
Images: Seedream 5 Pro (wavespeed generate image / edit image). Video: Seedance 2.5 (wavespeed generate video, edit video, extend video), with Kling 4.0, Wan 3.0 and MiniMax H3 (cheap, native audio, 768p) as alternatives. Everything else is one wavespeed models <query> away.
Support, privacy, terms
- Support: support@wavespeed.ai · GitHub issues
- Privacy policy: https://wavespeed.ai/static/privacy
- Terms of service: https://wavespeed.ai/static/terms
- Security: see SECURITY.md
License
MIT — same as the CLI and the MCP server.
WaveSpeedAI — AI image & video generation platform. Try it in the browser: Image generator · Video generator