great-az-tick-check-mcp — Model Context Protocol server for the UA Great Arizona Tick CheckA Model Context Protocol server that wraps the University of Arizona Cooperative Extension’s Great Arizona Tick Check program — Dr. Kathleen Walker’s lab in Forbes 410, Department of Entomology, UA Tucson — as a set of tools an LLM (Claude Desktop, Claude Code, Claude API agents, or any other MCP client) can call.
Built for the EpiHack Arizona 2026 Wildlife & Vector-Borne Diseases focus group and threaded into Scenario A (“Hiker mails in a tick”) of the end-to-end data flows.
Mock-by-default. The Great Arizona Tick Check is a mail-in program — there is no public REST API today. This server ships with an in-memory mock backend so the rest of the EpiHack stack can develop against it. The mock is good enough to demo the full create → status → species + pathogen-results flow in a single LLM session, and bad enough that you would never confuse it for the real lab. When a real backend ships, set
GATTC_BACKEND_URLin the environment and the client routes to it instead (see Pointing at a real backend).
| MCP tool | What it returns |
|---|---|
gattc_create_submission |
A new submission ID, the lab’s static mailing address, a placeholder mailing-label URL, a status URL, and the estimated turnaround in days. |
gattc_submission_status |
Current status (received → identifying → testing → complete); once complete it also returns species and pathogens_tested. |
gattc_species_identification_from_photo |
A low-confidence best-guess species + alternatives from a short list of AZ-relevant ticks. Always carries verify_with_lab: true. |
gattc_pathogens_screened |
Reference list of pathogens the Walker lab screens for, each with its ICD-10 code. |
gattc_mailing_label |
Placeholder mailing-label URL (PDF or PNG) plus the lab’s mailing address. |
Plus two MCP resources: gattc://mailing-address and
gattc://az-tick-species so an LLM can pull them as context without
having to call a tool.
The Great Arizona Tick Check is the closest thing Arizona has to a statewide participatory tick-surveillance program — exactly the pattern EpiHack is trying to scale across heat and vector-borne disease. Wrapping it as an MCP server lets an LLM:
The reference list returned by gattc_pathogens_screened. ICD-10
codes for Rickettsia rickettsii come from
schema/deep/standards.sql; the
others come from
schema/deep/pathogens.sql and
standard CDC tick-borne disease references.
| Pathogen | Disease | ICD-10 |
|---|---|---|
| Rickettsia rickettsii | Rocky Mountain spotted fever | A77.0 |
| Rickettsia parkeri | R. parkeri rickettsiosis | A77.8 |
| Borrelia burgdorferi | Lyme disease | A69.20 |
| Anaplasma phagocytophilum | Anaplasmosis | A77.49 |
| Babesia microti | Babesiosis | B60.0 |
| Ehrlichia chaffeensis | Ehrlichiosis | A77.40 |
The short list gattc_species_identification_from_photo chooses
from:
By default the server runs in mock mode — submissions live in a dict for the lifetime of the process. To swap in a real HTTP backend once the Walker lab (or a partner) ships one:
export GATTC_BACKEND_URL=https://great-arizona-tick-check.example/api
export GATTC_API_TOKEN=... # optional, if the real backend needs auth
The HTTP path in client.py
is intentionally a thin stub that raises NotImplementedError until
the request bodies + response parsing for the real API are filled in.
That way a misconfigured deployment (typo’d URL, wrong env) fails
loudly rather than silently masquerading as a working mock.
A .env.example template is included; copy to .env and source it.
This server doesn’t replace the real program — it shadows it. The
real mailing address (printed verbatim by gattc_create_submission
and the gattc://mailing-address resource) is:
Dr. Kathleen Walker Forbes 410, Department of Entomology P.O. Box 210036 University of Arizona Tucson, AZ 85721
See the Great Arizona Tick Check homepage and Help Us page for current submission instructions (a tick in a sealed bag, ideally frozen 1–2 days, with date + city/town + ZIP).
uv if you don’t have it.examples/claude_desktop_config.json
into your Claude Desktop config (~/Library/Application Support/Claude/claude_desktop_config.json
on macOS, %APPDATA%\Claude\claude_desktop_config.json on Windows).GATTC_BACKEND_URL empty to run in mock mode, or fill it in
for a real backend.cd mcp/great-az-tick-check-mcp
uv sync
uv run great-az-tick-check-mcp # stdio (default)
MCP_TRANSPORT=streamable-http uv run great-az-tick-check-mcp # HTTP
cd mcp/great-az-tick-check-mcp
uv run pytest
Synthetic-data tests exercise the full create → status → complete flow, the species-guess response shape, the deterministic Walker-lab mailing address text, mailing-label URL generation, and the pathogens-screened reference list. No live credentials and no network calls.
These are the deliberate stubs in this server. A future PR plugging the real backend in should at minimum:
client._HttpBackend.create() and .get() to POST/GET
against the real submission endpoint. The Submission dataclass
is the contract — whatever shape the real API returns has to map
into it.client._MockBackend.get() with whatever the real backend reports.client._label_url / client._status_url with whatever
signed-URL or routing scheme the real backend uses (override the
GATTC_LABEL_BASE and GATTC_STATUS_BASE env vars if just the
base changes).client.species_guess() (the
current implementation is deterministic and ignores the photo;
the verify_with_lab: true flag is the safety net).MIT, alongside the rest of epihack-2026.