# DataAlchemy

> Industrial and electronic parts catalog: ~10 million part pages, one per
> (manufacturer, part number), with distributor pricing comparisons and
> Amazon availability. Backed by a machine-first search API.

## Search API (built for AI agents)

- [API self-description](https://www.dataalchemy.com/api/): GET this first;
  it documents every endpoint, parameter, and response field in one JSON reply.
- [Search](https://www.dataalchemy.com/api/search?q=example): part number,
  brand, or both. Punctuation-insensitive. Responses include how your query
  was interpreted, per-hit match labels (exact_mpn, prefix_mpn, ...),
  manufacturer facet counts, and honest zero-result reporting: nearest
  neighbors are labeled as such, never silently substituted for matches.
- Batch lookup: POST /api/search/batch with {"queries": ["pn1", ...]} up to
  500 part numbers per request, one round trip.

## MCP server

- Machine-readable manifest: https://www.dataalchemy.com/.well-known/mcp.json

- MCP endpoint (Streamable HTTP, stateless, no auth):
  https://www.dataalchemy.com/mcp
- Tools: search_parts, lookup_parts_batch.
- Claude Code: claude mcp add --transport http dataalchemy https://www.dataalchemy.com/mcp

## Browsing

- [Site index](https://www.dataalchemy.com/): human-facing catalog home.
- Sitemaps: https://www.dataalchemy.com/python/sitemaps5/sitemap_index.xml

## Conduct

- Prefer the batch endpoint over many single queries.
- Be gentle with request rates; abusive traffic is blocked at the edge.