How AI agents connect (MCP)
How Claude and other AI agents work with your Toreca Cloud stock and purchases.
Toreca Cloud has an endpoint for AI agents. It follows MCP (Model Context Protocol), a shared standard, so any MCP-capable app such as Claude Code, Claude or Cursor can use it as is.
Once connected, you can simply ask your agent “Which shelf holds our QA card stock?” and it looks it up in Toreca Cloud. Within the access you grant, it can also move stock or record receipts for you.
Connection details
| Item | Value |
|---|---|
| MCP URL | https://api.toreca-cloud.com/mcp |
| Authentication | OAuth (sign in to Toreca Cloud in your browser and approve) |
| Connection lifetime | 12 hours from approval |
There is no API key to issue. Add the URL to your app and a browser opens on first use. Sign in to Toreca Cloud and pick a work role to finish connecting.
Built-in safeguards
- You pick a work role every time you connect. The role decides which operations the agent can see (work roles and connection lifetime).
- The agent never goes beyond the signed-in member’s permissions. Anything your role in the shop forbids is forbidden over MCP too.
- Every operation runs through the same logic as the console and is recorded in the audit log.
- Identity document images and full bank account numbers are never handed to MCP.