Recover assets from a page

With the MiroMiro tools connected (Claude, ChatGPT, Cursor, Codex) use the tools in the table. Without them (OpenClaw and other agents) call the matching REST endpoint: /v1/images, /v1/svg, /v1/videos, /v1/audio, /v1/lottie.

curl -s "https://miromiro.app/api/v1/svg?url=github.com" -H "Authorization: Bearer $MIROMIRO_API_KEY"

No key yet? The user gets one free at https://miromiro.app/api/onboarding (300 credits a month, no card). Ask them for it; never invent one. Images, SVGs, videos and audio cost 1 credit; Lottie 3.

One tool per asset kind. All of them take a url and return resolved URLs tagged by where they were found.

Ask	Tool
Icons, vector logos, anything SVG	extract_svgs
Images, photos, og:image, favicons	extract_assets
Video files and player embeds	extract_videos
Audio files, podcast enclosures	extract_audio
Lottie / animation JSON	extract_lottie

extract_svgs returns inline SVG markup sanitized and paste-ready, alongside external .svg file references — so an icon set can usually be lifted straight into a project.

The JavaScript caveat, which matters most here

The API reads static HTML by default. Media is the category most often mounted at runtime, so thin results are common and expected:

• Videos — most sites mount their player with JavaScript
• Lottie — most Lotties are runtime-injected
• Audio — JavaScript players are the norm

When any of these come back empty or suspiciously short, call again with render: true. That runs the page in a real browser so its JavaScript executes first (+5 credits). For video, audio and Lottie specifically, an empty first result is a reason to retry, not a reason to tell the user the page has none.

Check the note field on video and audio responses — it spells out the coverage caveat.

Reading video results

source: "embed" means a player page (YouTube, Vimeo, Wistia, Loom), not a downloadable file. Do not present an embed URL as if the user can download it — link it as an embed.

Result caps

Remote results are trimmed so a large page cannot flood the context, and every trimmed response states what was dropped. Two optional arguments loosen it:

• limit — raise the item cap, up to 200
• full_output: true — return complete long strings, such as entire inline SVG markup

Reach for limit when the user wants everything on an asset-heavy page, and full_output when a truncated SVG needs to be complete to be usable.

Before handing assets over

Assets on someone's site are usually theirs. Extracting for reference, analysis, or rebuilding your own version is ordinary work; redistributing their logo, photography, or video is a licensing question. Mention it when the user is clearly heading toward republishing rather than referencing.