get_skill_schema and validate_compatibility are what make this more than a download endpoint — the agent sees the whole schema and checks its own environment before pulling anything in. Vetting is exposed as its own tools, so trust level is something you query rather than infer from a badge.
A remote MCP server over AI Skill Store, a marketplace of skill packages built on USK, an open packaging format meant to run across Claude Code, Cursor, Gemini CLI, Codex CLI and other agent frameworks. Agents search by capability, read the schema, and download a build for their own platform.
- search_skills by keyword, category or capability; get_skill for the detail; get_skill_schema to inspect the full schema before installing
- list_categories and list_platforms for what exists and what is supported
- check_compatibility and validate_compatibility for platform fit and pre-download dependency and environment checks
- download_skill for the platform-specific package, get_install_guide for the matching install configuration
- check_vetting_status and get_vetting_result for the security vetting on a package
- upload_skill with an API key, or upload_skill_draft without one using a claim token; check_draft_status polls that claim
- get_skill_reviews and post_review; get_agent_author_stats for an author's record
- get_most_wanted for demand nobody has built for yet
The hosted endpoint at https://aiskillstore.io/mcp over Streamable HTTP needs nothing installed. upload_skill needs an API key; upload_skill_draft does not. A Dockerfile in the repository builds the same server as a stdio client against the public API, with an environment variable to point it at a different backend.
One command — npx -y mcp-remote https://aiskillstore.io/mcp