pricepilot-mcp
README.md
# PricePilot MCP Server
> **Status: retired (2026).** PricePilot ran as a live, public MCP server. It was listed in Anthropic's MCP directory, on Glama and Smithery, and in several other MCP directories, and all six tools were tested in real use. The hosted server and the Claude Desktop extension have since been retired, so the connection steps below will no longer work. The code stays public as a reference build; to run it yourself, use the local setup steps with your own database.
Free Amazon pricing intelligence for multi-channel CPG brands, exposed as a Model Context Protocol server. Six read-only tools over a point-in-time Amazon Buy Box snapshot across Grocery, Health & Beauty, Household, and Pet Supplies.
A free alternative to NielsenIQ / SPINS syndicated category data — accessible from any MCP client (Claude Desktop, Claude.ai, Cursor, Continue, agent frameworks).
## What it does
Given a price and a category, the server answers questions like:
- Where does this Amazon price sit against 100+ tracked competitors?
- Is the category trending up, stable, or falling over the last 30 days?
- What are the budget / midmarket / premium price bands?
- How do these N SKUs stack against each other on the shelf?
The server returns derived statistics only — percentile rank, Price Index, trend direction, tier bands. No raw competitor prices are exposed.
## Connect
### Hosted endpoint (retired)
```json
{
"mcpServers": {
"pricepilot": {
"url": "https://pricepilot-mcp.onrender.com/mcp"
}
}
}
```
No API key. Rate limited to 60 requests/minute, 1000/day.
### Claude Desktop extension (retired)
Use the official `.mcpb` extension at https://github.com/vantage-meridian-group/pricepilot-mcpb/releases/latest — drag-and-drop into Claude Desktop.
### Local stdio (for development)
```bash
pip install -e .
DATABASE_URL=postgresql://... python -m pricepilot_mcp
```
## Tools
| Tool | What it returns |
|---|---|
| `get_price_position` | Percentile rank, Price Index, position (Value / Parity / Premium) for a single price |
| `get_category_trend` | 30-day trend direction (Rising / Stable / Falling) with sample size |
| `get_category_overview` | Tier breakdown (budget / midmarket / premium bands) and median |
| `compare_products` | Per-SKU stack rank for a list of products against the category |
| `list_categories` | Available categories with product counts and trend — call this first |
| `server_status` | Health and data freshness (degraded if seed is >10 days old) |
All tools are `readOnlyHint=true`, `destructiveHint=false`, `openWorldHint=false`.
## Categories
A point-in-time snapshot from Amazon Buy Box scans (see `server_status` for the snapshot date):
- Grocery & Gourmet Food
- Health & Beauty
- Household
- Pet Supplies
## Architecture
- FastMCP with Streamable HTTP and stdio transports
- SQLAlchemy + PostgreSQL for benchmark snapshots (read-only from the server's perspective)
- Rate-limited per-consumer with daily and per-minute buckets
- Dockerfile included; deploys to any container host (production runs on Render)
## Run with Docker
```bash
docker build -t pricepilot-mcp .
docker run -p 8081:8081 -e DATABASE_URL=postgresql://... pricepilot-mcp
```
## Operator note
The MCP server is a free, category-level, point-in-time snapshot of PricePilot data. The paid per-SKU analysis platform is not currently available. More about the team at https://vantagemeridiangroup.com.
## License
MIT — see [LICENSE](LICENSE).
TDQS
A4.6/5.0
Scored across 6 tools
Disambiguation5/5
Each tool has a distinct purpose: listing categories, getting overview, trend, single price position, multi-product comparison, and server health. No overlap.
Naming Consistency4/5
Most tools use verb_noun pattern (list_categories, get_category_overview, etc.), but 'server_status' is a noun_noun exception. Overall consistent and readable.
Tool Count5/5
6 tools is well-scoped for a pricing intelligence server; each tool serves a clear role without being too few or too many.
Completeness5/5
Covers all essential operations: listing categories, getting overview/trend, single and multi-product price analysis, and server health. No obvious gaps.
Maintenance
ActivityMaintained
ResponsivenessNo issues