AI Toolkit
Connect your AI tools to iOrdertrack with the iordertrack-ai-toolkit Claude plugin.
The iOrdertrack AI Toolkit connects your AI tools to the iOrdertrack
platform. With the Toolkit you can build against iOrdertrack's documentation,
API schemas, and code validation, and read your store through the toolkit's
store_execute capability. The Toolkit ensures your agent works with
iOrdertrack correctly, rather than guessing at how things are implemented.
Install
claude plugin marketplace add iordertrack/claude-plugins
claude plugin install iordertrack-ai-toolkit@iordertrackThen provide your credentials (a personal access token) via environment variables:
export IORDERTRACK_API_TOKEN=iot_pat_xxxxxxxxxxxxxxxxxxxxxxxx
export IORDERTRACK_API_URL=https://api.iordertrack.com/api # optional; this is the defaultTools
| Tool | What it does |
|---|---|
search_docs | Search these developer docs and return relevant excerpts + links. |
get_api_schema | Return the OpenAPI schema for a resource or operation (params, shapes). |
validate_code | Validate a storefront section/theme config or an API request against our contracts. |
store_execute | Read your store: products, orders, inventory, sections, theme, sales report. |
store_write | Write your store: products, inventory, and storefront (theme / sections). Needs a *:write scope. |
What you can do today
store_execute reads your store — list and inspect products, orders, inventory,
sections, theme, and your sales summary (get_sales_report).
store_write makes changes using a token that carries the matching write scope:
create_product,upsert_inventory—catalog:write/inventory:writeupdate_storefront,update_theme,add_section,update_section,reorder_sections,delete_section—store:write
It's a separate tool from store_execute on purpose, so an agent never writes
when asked to read. See Managing products and
Managing your storefront for the endpoints, and
Authentication for how write scopes are enforced
(fail-closed: only blessed endpoints are reachable).
Why use it
Without the Toolkit, an AI agent guesses at endpoint names, payload shapes, and validation rules. With it, the agent reads the real schema, validates payloads before sending them, and searches authoritative docs — so it builds against iOrdertrack the way it actually works.