AI backlink outreach agent or catalogue basket? What the agent can do here
An agent can shop our catalogue for you, but it does not email webmasters, and it cannot spend your balance without you. This page lays out the split.

Outreach versus the shelf
Backlink outreach software sends emails to site owners and waits for replies. Our catalogue works the other way round. The sites are already listed, each with a price, and you pick rows from the 50,000+ in stock without a single cold email. Simple enough.
That difference changes what an agent is for. It does not write pitches here. It browses donors, compares rows against the instructions you gave it, and fills the basket, which suits resellers who maintain many client accounts and want a first draft of each order before a human looks.
An example, invented for the page. A reseller with six small clients writes one brief per client in a shared document: niche, language, a ceiling of $60 per row, a minimum traffic figure and two target URLs. Each Monday the agent drafts a basket per client. The reseller reads six baskets in twenty minutes, removes the rows that fail the scorecard, and pays the rest.
What the agent cannot do matters as much. It cannot know that a client changed its mind on Friday. Be realistic. It cannot see your agency's quiet history with a donor. And it will never be the one who answers when a client asks why a link is where it is. Keep that conversation yours, and keep a test order in your routine for any new donor type the agent proposes.
Be careful with vague briefs. An agent told to find good links will define good by whatever columns are easiest to compare, usually Domain Rating, and that is the trap we warn about in the row glossary. Name the traits you care about. Tell it what to skip, too.
Logs help. Ask the agent to write one line per row explaining why it chose that donor, then store those lines with the basket. When a client asks in January why a certain site was used in November, you read the line back instead of guessing, and the agent's reasoning becomes part of your paperwork.
Security deserves a sentence of its own. Treat the Bearer key like a password, keep one key per project, and never paste it into a shared document. Rotate it if a freelancer leaves. An agent with a key can fill baskets, and baskets are only safe while a human still decides what to pay.
Briefing the agent, one client at a time
An ai backlink outreach agent is only as careful as the brief it receives. Write one brief per client, in the same place every time, so the rules travel with the account. Boring is good. A vague line such as find good links gets you whatever is easiest to sort, and that is usually one number.
Treat the brief as the contract between you and your assistant. Keep it short enough to read in a minute, and specific enough that two people would pick the same rows from it.
The fields every brief needs
Niche, language, a ceiling per row, a minimum traffic figure, and the target URL and keyword for each page. Those five fields cover most of what a client cares about. Missing one is common. Leave one out and the agent fills the gap with its own guess.
Write them in the same order each time, so you can compare two briefs at a glance.
A ceiling per row, not per order
Set the ceiling on each row. A total ceiling lets the agent pick one expensive donor and fill the rest with weak ones. A row ceiling forces a spread. Say, for example, that no row should cost more than $60 on a small account.
Revisit the ceiling when the label changes, not every time a donor looks tempting.
What to tell it to skip
Name the donor kinds you refuse, such as sites with no byline, pages that are mostly link lists, or any language your client does not write in. Negative rules are often clearer than positive ones. Say no clearly. An agent can check them against each row without much judgement.
What the agent does, step by step
Connect over MCP or the REST API
ChatGPT and Claude can work over MCP. Developers can also order through the REST API with a Bearer key.
Give it written rules
Niche, language, a price ceiling, a minimum traffic figure, and the target URL and keyword for each client.
Let it browse and fill the basket
The agent can browse donors and add rows. Adding to the basket charges nothing.
You review and pay
Paying the basket is a separate step that a human approves. Balance is topped up in the dashboard, not through the API.

Reading the basket before you pay
A basket the agent builds is a draft, nothing more. Adding rows costs nothing, so you can read it at leisure, compare it with your own pick and throw out what does not fit. Drafts are cheap. Give the review a set time, so that a long list does not pass for a checked one.
Open the rows you are least sure of first. Check each against your scorecard, and mark any row that fails the niche or language test, since those mistakes are the hardest to explain later.
A focused read is a short sitting
Once the briefs are sharp, a basket for one client is usually a short sitting. Time goes on the rows that make you pause, not on the ones that obviously pass or fail.
If every basket takes an hour, go back to the brief before you blame the agent.
Rows you remove, and why
Keep a note of every row you drop from the agent's list and the reason for dropping it. Over a few weeks the notes show where the brief is too loose. Look for patterns. You can then tighten it without guessing.
An agent will fill a basket exactly as loosely as you brief it. Paste your five scorecard questions into its instructions, and ask it to explain each row it adds in one line.
Questions people ask about agent ordering
Can an ai backlink outreach agent send emails to site owners for me?
Not through this catalogue. The sites here are already listed with their prices, so there is no cold email step. An agent can browse the rows, compare them with your brief and fill a basket. Outreach software is a different tool with a different job.
Does the agent pay for the basket?
No. Paying is a separate step a human approves. Balance is topped up in the dashboard, and the API does not do top-ups. Keep it that way for every client account.
Which assistants can order this way?
ChatGPT and Claude can work over MCP. Developers can also order through the REST API with a Bearer key. Either route still ends with a basket you review before anything is paid.
Is one agent enough for many clients?
One agent with a separate brief per client works, provided each brief stays on its own sheet. Mixing two clients' rules in one brief is the quickest way to put the wrong donor in the wrong basket.
What if the agent suggests a donor type I have not tested?
Treat it as a new test. Order one cheap row from that kind of donor first, read the live article and log the result, before you build a larger bundle around it.
Who does what
| Task | Agent | You |
|---|---|---|
| Browse and compare donors | Yes | Review |
| Fill the basket | Yes | Review |
| Pay the basket | No | Yes |
| Top up the balance | No, dashboard only | Yes |
| Promise results to a client | Never | Never |
| Check live links in Ahrefs or Search Console | Can help | Final say |
Before you approve a basket
- Every row matches the client's niche and language.
- No single donor carries more than its share of the basket.
- Each row has a target URL and keyword taken from the brief.
- Each row has a one-line reason from the agent, and you agree with it.
- The total sits inside the client's budget and reserve.
- The list has been read twice before anything is paid.

Where a human still has to lead
Some decisions never move to the agent. Whether a client should buy at all, whether a donor deserves a second chance after a bad article, and what to say when a link disappears are all yours. The agent can prepare the paperwork for those choices. It cannot make them.
Keep one short note per client on what the agent may and may not decide. Read it before each new brief. A note like that is cheap to write and expensive to lose, because it records the boundary you set when you were calm.
Review those notes when a client changes its market, a donor changes its topic or the catalogue changes a column you rely on. A boundary that made sense earlier may need moving, and the note is where you move it.
When a client asks about a link the agent chose
Answer from the log, not from memory. The line the agent wrote for that row, together with your own note on it, tells you why it was picked.
If the honest answer is that nobody checked the row closely enough, say so and fix the brief. Admitting that is better than inventing a reason.