mcpField
E
epitaxy

epitaxy

Remote
drugs.crossgrain.xyzHealth0Updated 7w ago

About epitaxy

epitaxy is an MCP server for exploring US federal drug shortages and recalls alongside federal contracts. It links data by supplier while labeling match confidence and clarifying that a contract is not proof it supplies a specific drug.

Key features

  1. 01Drug shortage listings
  2. 02Drug recall listings
  3. 03Federal drug contract listings
  4. 04Supplier exposure matching
  5. 05Match confidence and freshness data

Use cases

  1. 01Check whether a supplier is associated with an active shortage
  2. 02Review recalls alongside federal contract records
  3. 03Inspect match confidence before using linked data

MCP support

Provides tools for MCP clients to query data freshness, drug shortages, recalls, federal drug contracts, and supplier exposure. Available as a remote server or through local stdio.

Configuration

Remote servers need no local install. Pick your client, then copy the config or install in one click.

Endpointhttps://drugs.crossgrain.xyz/v1/mcp
Copy install command
Claude CodeTerminal
Cursor.cursor/mcp.json
VS Code.vscode/mcp.json
Claude Desktopclaude_desktop_config.json
Windsurf~/.codeium/windsurf/mcp_config.json

Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.

Packages
pypiepitaxy-mcp0.1.4

Tools

5

This tool takes no parameters.

Details

From the project README

epitaxy

mcp-name: xyz.crossgrain/epitaxy

Federal drug shortages and recalls, joined to the federal contracts that buy those drugs. One call, no signup.

What it claims: this government supplier has an active FDA shortage or recall. What it does not claim: that a specific contract supplies a specific drug. Federal contract descriptions average 47 characters and almost never name the drug, so we say so instead of pretending otherwise.

Try it now, no key

curl https://drugs.crossgrain.xyz/v1/exposure?limit=5

One row of the answer, shortened:

{"exposure_id": "0026624dec7b6d9f",
 "award_id": "SPE2DP25F109W",
 "company_id": "a4fdd4a7530f4962",
 "company_name": "TEVA PHARMACEUTICALS USA, INC.",
 "company_key": "teva pharmaceuticals usa",
 "shortage_id": "022c478fb3af28c4",
 "confidence": "probable",
 "match_method": "company_name_normalized"}

MCP

Remote, nothing to install:

{ "mcpServers": { "epitaxy": { "url": "https://drugs.crossgrain.xyz/v1/mcp" } } }

Local, stdio:

{ "mcpServers": { "epitaxy": { "command": "uvx", "args": ["epitaxy-mcp"] } } }

Tools: get_data_freshness, list_drug_shortages, list_drug_recalls, list_federal_drug_contracts, get_supplier_exposure.

Read the confidence field, always

valuemeaning
exactmatch on a shared identifier, here an NDC inside FDA data
probablenormalized company name, or a drug name token in the contract text
weakanything softer, an indication only

Matching on both company and drug at once is not promoted to exact. Cross source joins are never exact, because FDA and USAspending share no identifier. Anyone claiming otherwise has a different source or is guessing.

Free tier

Works forever, no key, no signup. Returns at most 20 rows, without the evidence field, without weak matches, and with data delayed by 24 hours. Nothing is stored: no counters, no IP addresses. If a version that old does not exist yet, you get the oldest one available, so the free tier is never empty while data exists.

Addressing

These two addresses are the same segment and both work permanently:

https://drugs.crossgrain.xyz/v1/exposure
https://drugs.crossgrain.xyz/drugs/v1/exposure

Later segments arrive as /devices/v1/..., /food/v1/... and so on. A path whose segment is reserved but not live yet answers 404 with segment_not_live, it never returns another segment's data. An address that has been published is never withdrawn.

Paid

planpricecontents
Starter29 USD/mo10 000 calls
Pro99 USD/mo100 000 calls plus daily delta feed
x4020.004 USD/callUSDC on Base, no signup
Bulk dump199 USDone off, Stripe only

Sources and licence

  • openFDA, public domain, CC0 1.0 Universal
  • USAspending, public domain, code under CC0

Data is unvalidated and must not be used for medical decisions. This project is MIT licensed. Docs: https://drugs.crossgrain.xyz/docs

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.