Build it with the agent you already use.
Our docs are written for coding agents as much as for people — a machine-readable index at /llms.txt, every page mirrored as raw Markdown, a live OpenAPI spec. Hand them to Claude Code, Cursor, or whatever you run, tell it what you're building, and it writes working code against the API. Rather do it by hand? The quickstart is three curls.
Paste this into a fresh session:
We're adding an AI support agent to our product with Ingram Cloud. Read https://cloud.ingram.tech/llms.txt to learn the API, then build it: one assistant per signed-in customer, answering from our help center, and pausing for me before it issues a refund.
A few places to point it.
Each is one agent design, run as a private instance per person — with its own memory, its own tools, and a record of everything it does.
A private copilot per customer
One assistant for every user in your product — its own memory and connected accounts, with nothing crossing between them.
Learn howA support agent on your site
Answers from your help center, recognizes returning visitors, and hands the hard ones to a person. Signed-in or not.
Learn howAgents in Slack, WhatsApp, email
Deploy once. An inbound message wakes the right one; the reply returns to the same thread.
Learn howReal actions, gated by a human
Let an agent issue the refund or open the ticket — and stop for your sign-off on the calls that warrant it.
Learn howOr turn it around and expose your own product as an MCP server, so an agent in Claude or an IDE can drive it. The whole map is the ecosystem overview.
Spin up a project.
A project and a token, and you're a few minutes from your first reply — traced, costed, and replayable from the first one.