SaveAnything AI Agent Skill

Install it in your agent

The skill is a standard SKILL.md package that teaches any agent how to use SaveAnything, plus runnable examples in Bash, Python and Node.

1. 1

   Claude Code

   Unzip the skill into your skills folder, then just ask it to save a link.

   unzip saveanything-skill.zip -d ~/.claude/skills

2. 2

   Claude (claude.ai)

   Go to Settings → Capabilities → Skills → Upload skill, and pick the downloaded zip. Skills are available on paid plans.

3. 3

   ChatGPT & every other agent

   Paste the contents of SKILL.md into your custom GPT, system prompt, or agent instructions — it is plain markdown.

One endpoint, any agent

Claude, ChatGPT, or any assistant that can make an HTTP request can download media through SaveAnything. There is exactly one call to learn: post a public Instagram link to our search endpoint and you get back direct, ready-to-fetch download links. No API key, no OAuth dance.

Copy-paste usage

Send POST /api/search with the Instagram URL in q. The other three fields are fixed:

curl -X POST https://saveanything.app/api/search \
  -H "Content-Type: application/json" \
  -d '{
    "q": "https://www.instagram.com/reel/ABC123/",
    "t": "media",
    "lang": "en",
    "v": "v2"
  }'

{
  "status": "ok",
  "media": [
    {
      "type": "video",
      "url": "/api/download?url=...&filename=SaveAnything.app_reel.mp4",
      "thumb": "/api/img?url=...",
      "filename": "SaveAnything.app_reel.mp4"
    }
  ]
}

• media[].url and media[].thumb are site-relative — prefix them with https://saveanything.app before fetching.
• Download links are time-limited (about an hour) — fetch them right after resolving, don't cache them.
• A non-okstatus means the link couldn't be resolved (private account, deleted post, or not an Instagram URL).