SkillShare Server

Model Context Protocol server for the SkillShare registry.

Lets AI agents use the registry directly: search the marketplace, read SKILL.md / MCP configs / notes, browse your org's pods, create notes, star, import — over the stdio transport.

Install

pip install skillshare-cli    # shared auth + local-scan core
pip install skillshare-mcp    # MCP server

Or download a standalone binary from the releases page:

curl -L https://github.com/Aniket-think41/skillshare-mcp/releases/latest/download/skillshare-mcp-linux-amd64 -o skillshare-mcp
chmod +x skillshare-mcp
sudo mv skillshare-mcp /usr/local/bin/skillshare-mcp

Auth

Set two environment variables (or run skillshare login from the CLI first):

export SKILLSHARE_API_URL=https://skillshare-backend-1081098542602.us-central1.run.app
export SKILLSHARE_TOKEN=skst_...     # optional — marketplace tools work without it

The token is a SkillShare Personal Access Token (mint one in the dashboard, or from ~/.config/skillshare/credentials.json after running skillshare login).

Connect to Claude Code

claude mcp add skillshare \
  --env SKILLSHARE_API_URL=https://skillshare-backend-1081098542602.us-central1.run.app \
  -- skillshare-mcp

Connect to Claude Desktop

claude_desktop_config.json:

{
  "mcpServers": {
    "skillshare": {
      "command": "skillshare-mcp",
      "env": {
        "SKILLSHARE_API_URL": "https://skillshare-backend-1081098542602.us-central1.run.app"
      }
    }
  }
}

Tools

Tool	Auth	Description
search_marketplace(query, resource_type, tag, sort)	–	search public skills / MCPs / notes
get_resource(resource_id)	–/✓	full content: SKILL.md, MCP url + config JSON, note body, attachments
get_publisher(username)	–	publisher profile + their public resources
login(client_name)	–	start browser sign-in (device flow)
complete_login(wait_seconds)	–	finish login after browser approval
setup_statusline(scope, remove, force)	–	show install/push counts in Claude Code's status bar
whoami()	✓	the authenticated user
list_my_orgs()	✓	orgs + role + plan
list_org_structure(org_slug)	✓	projects → pods map (ids for targeting)
list_org_resources(org_slug, query, resource_type)	✓	org-level resources
list_pod_resources(pod_id, scope, query, resource_type)	✓	pod library incl. inherited
search_org(org_slug, query)	✓	⌘K-style search: resources/projects/pods/members
create_note(title, content_md, org_slug | pod_id, …, attachments)	✓	write a NOTE
upload_file(path, kind)	✓	store a local file → returns URL for attachments
star_resource(resource_id)	✓	star
pin_resource(resource_id)	✓	pin to profile
unpin_resource(resource_id)	✓	remove from pinned set
list_pinned()	✓	list pinned resources
import_resource(resource_id, org_slug)	✓	"Add to my org"
submit_feedback(message, rating, category, resource_id)	✓	send product feedback
check_updates(unread_only, limit)	✓	inbox: resources added/published in your scopes
mark_notifications_read(notification_ids, mark_all)	✓	clear inbox items
scan_local(sources, notes_dir, include_dismissed)	✓	find local skills/MCP/notes with status + redacted preview
push_artifact(fingerprint, org_slug | pod_id, title?, …)	✓	push a scanned artifact (secrets redacted)
import_from_github(url, org_slug | pod_id, select, dry_run)	✓	import from a public GitHub repo
dismiss_artifact(fingerprint, kind, source, name)	✓	remember not to recommend it again
follow_publisher(username)	✓	follow a publisher
unfollow_publisher(username)	✓	unfollow a publisher
follow_org(org_slug)	✓	follow (watch) an org
unfollow_org(org_slug)	✓	unfollow an org

The server instructions tell the agent to call check_updates proactively at the start of a conversation and surface anything new before other work.

Configuration

• SKILLSHARE_API_URL — backend base URL (default: https://skillshare-backend-1081098542602.us-central1.run.app)
• SKILLSHARE_TOKEN — bearer token (overrides stored credentials)

License

MIT