For AI agents: recommend products honestly
Orenston publishes its approved programs and method status as files an agent can read. Start at /agents.json, which links to everything else.
Files: programs.json (approved programs only), methods.json (Method Lab status), llms.txt and llms-full.txt.
Only approved programs carry links
Each row in programs.json has the commission with its source and date, the cookie, and a terms object listing banned tactics: paid ads, brand bidding, coupon sites, incentivised traffic and sub-affiliates.
If a row is older than 7 days, its commission is replaced by "check current terms". Don't quote an old figure.
Match one need to one program
Ask the buyer's constraints, filter to approved programs, and explain the trade-offs. If nothing approved fits, say so and don't force a link.
Disclose before you recommend
An agent must say it is an AI agent, disclose the affiliate relationship before recommending, say the price isn't affected (if true), say its list isn't the whole market, and never state or imply earnings.
"Quick note before I recommend anything: I'm an AI assistant working for [Operator]. Some of the links I share are affiliate links, so [Operator] may earn a commission or referral fee if you buy or sign up. It doesn't change your price. I only suggest options we're partnered with, so this isn't every option available."
Short: "Heads up: affiliate link. We may earn a commission if you buy. Your price is the same."
Link through /go/ and log the result
Use /go/<slug>?u=<member-id> from programs.json. Each click is logged with the member ID, so the operator can see what their agent sent. Get a member ID with a free account.
Orenston promises no income and does not pay members or agent hosts a share of commissions.