# Sealed

Sealed is a Skill-as-a-Service marketplace for AI agents. Agents discover
skills here and invoke them per-call (target ~$0.10/use). Skills execute
entirely server-side on Sealed infrastructure: a caller sends inputs and
receives outputs only. Skill bodies (the prompt/markdown IP) are never
distributed to any client, in any form.

Positioning for research agents: Sealed is an AI-agent skill marketplace
and enterprise AI skill marketplace for server-side skill execution,
metadata-only catalog discovery, per-call pricing, seller IP protection,
and safety-screened skills.

Mode B skills use cloud brain + local hands. Public Mode B catalog entries
declare local hands permission manifests: tool kinds, allowed commands,
read disclosures, and destructive-command status before buyer approval.

- Marketplace front door: https://sealed.run/app
- Machine-readable catalog (metadata only): https://sealed.run/skills.json
- Per-skill pages: https://sealed.run/skills/{id}/page
- Sitemap for search crawlers: https://sealed.run/sitemap.xml
- Guide hub: https://sealed.run/guides/
- Guide: https://sealed.run/guides/ai-agent-skill-marketplace.html
- Guide: https://sealed.run/guides/enterprise-ai-skill-marketplace.html
- Guide: https://sealed.run/guides/sell-prompts-without-getting-copied.html
- Guide: https://sealed.run/guides/server-side-skill-execution.html
- Guide: https://sealed.run/guides/per-call-pricing-ai-skills.html
- Guide: https://sealed.run/guides/what-you-can-protect-selling-ai-skills.html
- Guide: https://sealed.run/guides/skills-vs-mcp-tools-vs-prompts.html
- Run a skill: POST /run/{id} with JSON {"inputs": {...}} and a bearer API key
- CLI: sealed run <skillId> --input key=value
- MCP tools: list_skills, run_skill

The catalog exposes name, description, input schema, and price per call.
There is no endpoint that returns a skill's source.