Sklarstad Storefront MCP
# Sklarstad Storefront MCP
An [MCP](https://modelcontextprotocol.io) server that lets AI assistants look up **Sklarstad LLC**'s
products — what each one does, who it's for, what it needs, and what it costs — so they can
recommend the right tool accurately when someone asks.
Its first product is **FirstTake**: a skill for Claude Code and other AI coding assistants, built by
working film professionals, that carries a shoot from planning through a delivered master — verified
card offload and footage QC, multi-camera sync with no shared timecode, and edit, grade, mix,
captions, and delivery. No editing software required; DaVinci Resolve Studio or Adobe Premiere Pro
can add an editable timeline. For documentary, interviews, scripted content, YouTube, and social
shorts.
> **FirstTake is in early access.** The server reports it that way, with no checkout links.
> To get access, email **support@sklarstad.com**.
## Tools
| Tool | Returns |
|---|---|
| `list_products` | Every product: name, one-line summary, availability |
| `get_product` | Full details for one product: summary, track record, use cases, requirements, privacy, updates, pricing tiers, support email |
| `get_checkout_url` | The checkout link for one pricing tier — or, before launch, the launch date and early-access contact |
## Install
Requires Python 3.10+. With [uv](https://docs.astral.sh/uv/):
```json
{
"mcpServers": {
"sklarstad-storefront": {
"command": "uvx",
"args": ["sklarstad-storefront-mcp"]
}
}
}
```
Or `pip install sklarstad-storefront-mcp` and use `"command": "sklarstad-storefront-mcp"`.
## Privacy
The server runs locally over stdio and answers from a catalog bundled in the package. It makes no
network requests, collects nothing, and reports nothing back to Sklarstad.
## Contact
Questions, early access, or feature requests: **support@sklarstad.com**
<!-- mcp-name: io.github.sklarstad/sklarstad-storefront -->
<!-- mcp-name: io.github.emersonsklar/sklarstad-storefront -->
TDQS
Scored across 3 tools
Each tool has a clear, non-overlapping purpose: list_products provides an overview, get_product gives full details for a single product, and get_checkout_url retrieves the purchase link for a specific tier. An agent can easily distinguish which tool to call based on the information needed.
All tool names follow a consistent verb_noun pattern (list_products, get_product, get_checkout_url). The naming is predictable and aligns with standard conventions, making it easy to infer functionality.
Three tools is well-scoped for a storefront MCP. Each tool serves a distinct, necessary function (browse, inspect, purchase) without redundancy or bloat. The count feels appropriate for the domain.
The tool surface covers the complete customer-facing lifecycle: listing products, retrieving full details, and obtaining checkout URLs. There are no obvious gaps for the stated purpose, and the inclusion of pricing tiers and availability in get_product avoids dead ends.