Lightweight MCP proxy for Claude Skills (auto-downloads backend)
Until 2026-09-26 this record combined several servers published from one repository. The others now have records of their own: claude-skills-mcp-backend, iflow-mcp_k-dense-ai-claude-skills-mcp. Checks from before that date include them.
Wellknown found it in public sources; nobody has proven control of it yet. Claiming takes one click if the repository is under your GitHub account, or a small file on your domain otherwise. Verified owners get the badge, 15-minute checks, status alerts, edits that outrank crawled data, and a ranking boost.
Agents can do it too: POST https://wellknown.network/api/v1/claims with {"agent":"k-dense-ai-claude-skills-mcp","method":"well_known_file"} — machine-readable steps at claim.json, guide at /docs/claim.
Everything here was measured by our prober or read from a registry. Nothing is self-reported.
Attributed to the source that supplied each field. Treated as claims, not facts.
# Claude Skills MCP Frontend Lightweight MCP proxy for Claude Skills that auto-downloads the heavy backend on demand. ## Overview This is the frontend component of the Claude Skills MCP system. It's a lightweight proxy (~15 MB) that: - Starts instantly (<5 seconds) - Auto-downloads the backend when first needed - Acts as MCP server (stdio) for Cursor - Acts as MCP client (HTTP) for the backend - Returns tool schemas immediately (no backend wait needed) ## Installation ```bash # Via uvx (recommended for Cursor) uvx claude-skills-mcp # Via uv tool (persistent install) uv tool install claude-skills-mcp # Via pip pip install claude-skills-mcp ``` ## Usage with Cursor Add to your Cursor MCP settings (`~/.cursor/mcp.json`): ```json { "mcpServers": { "claude-skills": { "command": "uvx", "args": ["claude-skills-mcp"] } } } ``` Restart Cursor and the skills will be available! ### First Run Behavior On first run, the frontend will: 1. Start immediately (~5 seconds) ✅ **Cursor timeout satisfied!** 2. Return tool schemas to Cursor (instant) 3. Download backend in background (~250 MB, 60-120 seconds) 4. When you first use a tool, you'll see "Loading backend..." 5. Once backend ready, all tools work normally **Subsequent runs**: Fast! Backend is already installed. ## Configuration The frontend forwards all arguments to the backend: ```bash # Custom configuration uvx claude-skills-mcp --config my-config.json # Verbose logging uvx claude-skills-mcp --verbose # Custom backend port (advanced) uvx claude-skills-mcp --port 9000 ``` ## Remote Backend (Future) ```bash # Connect to hosted backend instead of local uvx claude-skills-mcp --remote https://skills.k-dense.ai/mcp ``` **Note**: Remote backend support coming in v1.1.0 ## How It Works ``` Cursor → Frontend (stdio, ~15 MB) ↓ list_tools() → Returns hardcoded schemas INSTANTLY ✅ ↓ [Backend downloads in background...] ↓ call_too…
Mapped onto the structured taxonomy from declared text and observed tool names. Confidence shown for derived entries.
Every source is kept verbatim. Field changes are logged as events.