Ask in plain language
Stock questions answered inside Gemini from live numbers, no dashboard switching.
Google's Gemini tooling speaks MCP too. Add Inventory Inventory as an MCP server in Gemini CLI or a Gemini-based agent, sign in with Google, and ask about live stock or record a sale from the terminal or your own agent.

Inventory Inventory publishes an MCP server — the open Model Context Protocol that Google's Gemini and the other major assistants use to reach outside systems. In Gemini CLI, add the server to your settings with the server URL (an "httpUrl" MCP server entry); the sign-in window opens in your browser. Gemini-based agents built with Google's agent tooling connect the same way.
Once connected, Gemini has six tools: list inventory, get an item, low stock, expiring soon, inventory summary, and record a sale. Reads return only items you marked “Include in API”, with quantity, price, cost, supplier, reorder level, expiry batches and location. A sale recorded from Gemini is transactional, never goes below zero, deducts the soonest-expiring batch first, lands in your activity log attributed to Gemini, and triggers alerts, auto-reorder and Square or Shopify sync like any other sale.
Access is OAuth 2.1 with PKCE: a consent screen names the app and its return address, tokens expire in an hour and rotate, and every connection is listed under Connected AI apps on your profile with a one-click revoke. No AI runs on our side for MCP — the model is the Gemini you already use.
Stock questions answered inside Gemini from live numbers, no dashboard switching.
Small on purpose: few context tokens, nothing to confuse the model.
record_sale is flagged destructive, so Gemini confirms first. FEFO on batches, always logged.
Google sign-in, consent screen, expiring tokens, revoke from your profile.
Per-item “Include in API” controls visibility — the same rule as the REST API.
The same server works in Claude, ChatGPT, Cursor, Gemini tooling and agents you build.
The best inventory interface is the one you are already typing into.
In Gemini CLI, add the server to your settings with the server URL (an "httpUrl" MCP server entry); the sign-in window opens in your browser. Gemini-based agents built with Google's agent tooling connect the same way.
A window from inventoryinventory.com asks you to sign in with Google and approve read access, and optionally recording sales.
Low stock, expiries, values, a specific SKU — or record a sale. Revoke any time from Profile → Connected AI apps.
In Gemini CLI, add an MCP server entry with the URL https://inventoryinventory.com/api/mcp to your settings (the mcpServers section, using the HTTP URL form). On the first call Gemini opens a browser window to sign in with Google and approve access. Agents built on Google's Gemini agent tooling with MCP support connect the same way. Wherever Google adds MCP connectors to the Gemini app, the same URL applies — the server is standard MCP.
The same six tools as every other assistant: list inventory, get an item, low stock, expiring soon, inventory summary, and record a sale if approved. Gemini reads live numbers and confirms before deducting stock.
Yes, and they complement each other. The built-in AI Assistant and AI Analyst run inside the app with no setup and use Google's Gemini models on our side. The MCP connector lets a Gemini you run yourself — in the CLI or your own agent — reach the same stock. No AI cost lands on your Inventory Inventory account for MCP.
Access uses OAuth 2.1 with PKCE, the standard Google and the major banks use for third-party apps. Tokens expire in an hour, refresh tokens rotate on every use, and any connection can be revoked from your profile. Gemini only sees items you exposed, and the one write tool must confirm with you and can never take stock below zero.
MCP access is part of the Pro plan at $10 per month, with the AI features, imports, purchasing, store integrations and the REST API. Free accounts can connect, but tool calls are declined with an upgrade link.
Start free with 50 items — no credit card, no trial countdown. Import your spreadsheet and you have a working system in minutes.