Your AI Agent Builds Backlinks Now.

Plug the agent in. Links flow out. Rankings climb.

What you're watching: the real session — 116 gap donors found for aiprofitboardroom.com, two placements picked, $225 total, live marketplace data.

Ask for pages. Get pages.

You describe what you want, and it comes back with actual pages — not domains.

What the ask looks like inside Claude.

Then you ask who links to your competitors — and not you.

One question shows you the gap.

Then you say: place the order. And it does.

You never open the platform.

The receipt: both orders live on the Insert.Link Orders dashboard — aijourn.com → aiprofitboardroom.com (anchor "AI automation community", $135) and coruzant.com → aiprofitboardroom.com (anchor "AI Profit Boardroom", $90). Status: In progress — the links go live when the publishers place them.

The Clicking Problem.

I've been doing SEO for years, and the part that ate my time was never strategy.

It was the clicking.

The part this removes: the clicking, not the strategy.

Four things, in order.

MCP is a plug and a socket.

Your AI agent is the plug.

The tool is the socket.

A custom connection for every single tool.

Different shape for every socket — slow to build, and it broke a lot.

Every agent-to-tool pair needed its own build — and every update snapped it.

One standard shape. Everything connects.

Any tool that supports it connects to any agent that supports it.

Guessing from old data vs reading the live database.

That gap is the whole thing.

Marketplaces are old news.

Link building marketplaces have been around for over a decade.

The marketplace isn't the news. The socket is.

The only marketplace in this category with an MCP integration.

The marketplace stops being a website you visit, and becomes a tool your agent uses.

Three things change — and the second one is the one nobody talks about.

A domain has 3,000 articles. Maybe 4 are about your topic.

Most link tools stop at the domain — Insert.Link indexed the pages.

Same site. Totally different result.

Google reads the page the link sits on — and the words around it.

The free research tools became agent tools.

Five lookups, free — no limits on results, no per-seat charge.

On their own, they're useful. Connected to an agent, they become something else.

Four tools, back to back, without you.

Manually that's an afternoon — through the agent it's one instruction.

Ten thousand domains. One check.

Any old list — a competitor's profile, a scrape, a dusty spreadsheet — answered in a batch.

That used to be manual checking, one by one. Now it's a batch.

One workspace per client. Each with its own spend limit.

Orders, budgets and history stay separate.

Three levels of automating this. You don't need to jump to the end.

Sign up, paste the connection link. That's the setup.

Then you talk to it like a research assistant with the whole marketplace open in front of them.

What you're watching: the real connection, start to finish — the MCP tab on app.insert.link, the Server URL, the one command in your agent, the Authorize click, and both agents live on the Active Sessions list.

You're not automating yet — you're removing the clicking.

Shortlist in seconds. Pick two.

Write your rules down once. The agent follows them every time.

This is what separates people who use AI from people who get compounding results from it.

---
name: link-building-rules
description: My link building standard. Read IN FULL before touching
  the Insert.Link MCP. These rules outrank speed. If a rule blocks an
  order, the order does not happen.
---

# Link Building Rules — the standard that doesn't drift

## 0. Before anything
1. Confirm which client workspace this session is for.
2. Load that workspace's spend limit. Never assume it.
3. Always run the Backlink Gap first — me + top 3 competitors.
   The gap report decides the week's targets, not my mood.
4. Pull the Anchor Text Lookup on our main money phrases so we
   know what anchors already point at us BEFORE adding more.

## 1. What counts as a target page
- Only pages actually about the topic. Never category pages,
  never tag archives, never "top 100 everything" roundups.
- The page's headline must name our niche or a sibling niche.
- Reject any page where our link would sit next to casino,
  pharma, or crypto-gambling anchors. No exceptions.
- Prefer pages already ranking for something (any position
  in the top 50 counts as a signal).

## 2. Site filters
- Only these niches: [marketing, AI, SaaS, business, ecommerce]
- Skip any site whose link profile looks off:
  - referring domains falling for 3+ months
  - anchor profile >40% exact-match commercial
  - outbound links per page >25
- Metrics are a tiebreaker, never the reason. A mid-metric page
  about our exact topic beats a big domain with an off-topic slot.

## 3. Anchors
- Always vary the anchor text. Rotate:
  brand name (40%), naked URL (20%), natural sentence
  fragment (25%), partial match (10%), exact match (5% max).
- Check the live anchor mix with Anchor Text Lookup before
  every batch. If exact-match is creeping up, order zero
  exact-match anchors this round.

## 4. Pace
- Max [N] new placements per site per month.
- Max [N] total new links per week per client.
- New client with a small profile? Start at half pace for
  the first month. Going 4 → 80 in a week is how you get noticed
  for the wrong reason.

## 5. Delivery
- Client needs speed → prefer the Fast Delivery list
  (placements can go live in up to 3 days).
- Otherwise optimise for relevance over speed, every time.

## 6. Budget
- Never exceed the workspace spend limit. Not by $1.
- If the shortlist total is over budget, cut the LEAST relevant
  pages first, never the cheapest.
- Log every order: page URL, price, anchor, date, workspace.

## 7. Output format
- Present every batch as a shortlist table BEFORE ordering:
  page URL · topic match · metrics · price · proposed anchor.
- Wait for my approval on the shortlist. I approve or I don't.
- After approval, place orders and report back: what was
  ordered, spend used, budget remaining.

## 8. Hard stops — ask me, never guess
- A site asks for content we didn't write → stop, ask.
- A price looks wrong by 2x either way → stop, ask.
- Anything about the client's brand positioning → stop, ask.

# A human skips steps at 6pm on a Friday. A written rule doesn't.

The full rules file — my agent reads this before it touches the marketplace. Hover to pause.

A human skips steps at 6pm on a Friday. A written rule doesn't.

Five clients? Five workspaces, five spend limits, one set of rules — the same standard across all of them.

Agents that share one memory — and now they can act.

This is what I run: an agent operating system where every agent knows what the others did.

Recording 1: Hermes inside my Agent OS runs the backlink gap through the Insert.Link MCP — live donors, prices, and its own recommendation ("portotheme.com is the standout value — DR 90 for $90").

Recording 2: same Hermes tab, the anchor lookup — 146 domains with live "AI SEO" anchors, streamed back with DR and prices.

The competitor agent spots it. The link agent acts on it.

The agents do the finding. You do the deciding.

You don't need Claude specifically. Any MCP agent works.

The tool underneath matters far less than the rules you give it.

Link building fails for three reasons — none of them are the tool.

Get this wrong and the automation just helps you make mistakes faster.

The number looks good in a spreadsheet. The link does very little.

This is exactly why page-level search matters more than domain-level.

Real links don't all say the same thing.

If every link uses the same exact phrase, the pattern is obvious.

The Anchor Text Lookup shows what already points at you — before you add more of the same.

Four links to eighty in a week doesn't look earned.

An agent will happily do all three wrong things at speed — that's why the rules file is the guardrail.

AI answers are built from sources.

Which pages get pulled depends on which pages are visible, linked, and clearly about the topic.

A link on a relevant page works twice now.

I run the same approach across my five sites — occupy several spots on page one at once, so whichever source an AI pulls from, it finds me.

Placements on relevant pages are one part of that — and the finding part is now fast.

A year ago this was a project. Now it's day one.

Insert.Link is first in this category. They won't be last.

The advantage isn't the platform. It's you, running agents.

Monday. Wednesday. Friday. None of it is complicated.

All of it used to take hours — and the agent isn't guessing, it's reading live data.

When a client asks what happened last month — it's in one place.

Three things worth keeping.

Run the backlink gap before you do anything else.

Insert.Link is free to sign up, and the five lookup tools cost nothing.