Linkagent

The agent for internal links

An AI agent that owns your internal linking

Tools give you suggestions. An agent closes the loop: crawl, score, review, ship, re-crawl. Linkagent runs that cycle on a schedule, so every new post gets linked into your site without anyone remembering to do it.

Free scan, no account needed. Takes about 20 seconds.

01

The agent crawls

On every scheduled run it re-reads your sitemap, picks up new and changed pages, and refreshes the topic model and link graph: orphans, clusters, and click depth included.

02

The agent proposes

New opportunities are scored and queued for review. Your past approvals and rejections persist, so the agent never re-asks a question you already answered.

03

The agent ships

Approved links go live through the embed script, WordPress plugin, or API, and stay live across re-crawls. Publish a post; it gets woven into your site automatically.

Why internal linking needs an agent, not another audit

Internal linking is not a project; it is a maintenance loop. Every post you publish arrives as an orphan. Every new cluster shifts which pages should point where. A one-time audit, however thorough, starts decaying the day you run it. This is exactly the shape of problem agents are for: repetitive, continuous, rule-bound work where a human should set policy and review judgment calls, not push every button.

Linkagent splits the work accordingly. The agent handles everything mechanical: crawling on schedule (daily, weekly, or monthly), rebuilding the content model, scoring page pairs, checking link budgets, and injecting approved links. You handle policy: approve or reject suggestions, and the agent treats those decisions as permanent instructions. Reject a link once and it never comes back; approve one and it survives every re-crawl, redesign, and CMS migration, because links live in the injection layer, not hardcoded in your content.

The judgment layer uses AI where AI is actually good (assessing whether a link helps a reader in a specific sentence) under hard constraints: anchors must already exist in your copy, budgets cap links per page, and nothing ships unreviewed. If you want the technique without the loop, read about our AI internal linking approach; if you want to see the agent's raw material, run the free internal link checker on your site.

Agent behavior you stay in charge of

Scheduled autonomy

Choose the re-crawl cadence per site. The agent works between your visits and has a review queue waiting when you return; it never publishes silently.

Decisions as memory

Approvals and rejections are the agent's long-term memory. It compounds your editorial judgment instead of resetting it on every run.

Deterministic guardrails

Link budgets, anchor diversity, and inbound caps are enforced in code, not left to model temperature. The AI proposes; the rules dispose.

Self-hostable agent

AGPL open source. Run the whole loop on your own infrastructure and point it at your own Claude key, or none at all.

Frequently asked questions

What is an internal linking AI agent?
Software that runs the complete internal linking cycle autonomously: crawl the site, model the content, score opportunities, collect human approval, ship the links, and repeat on a schedule. It differs from an internal linking tool the way a thermostat differs from a thermometer: it doesn't just measure, it acts.
Does the agent publish links without my approval?
No. Every suggestion sits in a review queue until you approve it. What the agent automates is everything around that decision (finding, scoring, injecting, and maintaining), so approval is the only thing left for a human to do.
What happens when I publish a new post?
On the next scheduled crawl the agent finds it, adds it to the topic model, and proposes links both directions: from established pages into the new post (so it isn't an orphan) and from the new post out to related pages. That is the moment most sites miss, and it is the agent's whole reason to exist.
Can I run the agent on my own servers?
Yes. The entire agent is open source under AGPL-3.0. Self-host it, schedule it with your own cron, and keep every byte of your content graph in-house.

Keep exploring

See what your site is missing

Run a free scan. No account needed, results in about 20 seconds.

Free scan, no account needed. Takes about 20 seconds.