MCP Server
AI agent integration via Model Context Protocol
Overview
The SkillPort MCP Server allows AI agents (Claude, OpenClaw, and other MCP-compatible clients) to programmatically search, evaluate, and install skills from the marketplace. It uses the Model Context Protocol (MCP) with stdio transport.
Setup
Add to your MCP client configuration:
Set the SKILLPORT_API_URL environment variable to point to a custom API server (defaults to https://api.skillport.market).
Tools
search_skills
Search the SkillPort marketplace for skills matching a query.
query (string), category? (string), max_price? (number), min_rating? (number), os? (string)get_skill_details
Get full details of a skill including permissions, danger flags, and install command.
skill_id (string)check_skill_safety
Evaluate the safety of a skill. Returns risk score, danger flags, and a recommendation.
skill_id (string)generate_install_command
Generate the CLI install command for a skill, including any required flags.
skill_id (string), version? (string)suggest_skills
Get skill recommendations based on a task description. Returns ranked matches from the marketplace.
task_description (string), platform? (string), limit? (number)inspect_skill
Inspect a skill's metadata, security info, inputs/outputs, and quality score.
skill_id (string)plan_install
Preview what an install will do before applying. Returns files, security summary, and rollback command.
skill_id (string), scope? ("project" | "global")check_policy
Check whether an action is allowed by the current .skillportrc policy. Returns allowed/blocked with reasons.
skill_id (string), policy_path? (string)Resources
ssp://catalog/popularTop 20 most popular skills by download count.
ssp://catalog/new20 most recently published skills.