JavaScript is off. The country switcher and the demos need it; everything else on this page works.
Open an account before 31 December 2026 and pay €1.95 instead of €2.95 per site per month, for life. The conditions
WebCrew tokens
Point Claude or ChatGPT at a WordPress site and you pay, in your own tokens, for every attempt it makes to work out what it is looking at. WebCrew already knows your sites and how each plugin stores its data, so it answers in one go. Every site gets 1,000 WebCrew tokens a month for that.
1,000 tokens per site every month, on Free and on Standard. No card.
The problem with your own tokens
Three situations, minute by minute, and what happens once WebCrew sits between your tool and your sites.
Every question about my sites burns tokens while the model figures out what it is looking at.
Tokens pay for the answer, not for the trial and error.
To make it useful I would have to give it a key to a live client site.
Reading is instant. Every write waits for a person.
It does not know how Elementor stores a layout or where ACF keeps a field.
Plugin by plugin, layout by layout: the number above counts what it can handle for you today.
The unit
A WebCrew token is our own unit; it is not a model token and it does not move with a provider's price list. An action from your own AI tool over MCP costs a fixed, small number. An action by the assistant, where WebCrew's model does the thinking, costs more, and you see that number before you approve.
What a token buys
The exact number is on every approval before you say go. Here is the shape of it.
| Task | Rate | What your own tokens would be doing instead |
|---|---|---|
| A question about your sites ("which sites run PHP 8.1?") | From 1 token, over MCP | Nothing. Your tool cannot see your sites without WebCrew. |
| A summary of what changed on one site this week | A few tokens, over MCP | Reading logs it does not have access to. |
| A draft that rewrites one page's text or settings | Assistant rate, shown in the log | Working out where the text lives, then writing to the live site with no restore point. |
| A draft across many items (prices on 40 products, metas on 60 pages) | Assistant rate, one number for the whole batch | Forty separate attempts, each one paid for, each one unreviewed. |
| Monitoring, security scan, updates, restore points | 0, always | In the site price. Never tokens. |
Your own AI tools use tokens today. The assistant inside wp-admin Coming soon.
For agencies
Like hosting or domains: WebCrew sells tokens to you, you set the price in your care plan. The client sees your logo on the assistant and your line on the invoice. A cap per site limits what each retainer can spend.
Questions
The 1,000 included every month reset on the first of the month; what you did not use is gone. Tokens you buy stay on your balance until you use them. Bonus tokens on top-ups for accounts opened before 31 December 2026 behave like bought tokens.
Keep it, and connect it. Your tool asks WebCrew over MCP and gets an answer from the register instead of guessing from a page. You pay tokens for the answer rather than tokens for the attempts, and every write takes a restore point first instead of landing on the live site with no way back.
A purpose-built layer of skills, tools and site context on top of frontier models: how each plugin stores its data, what a safe update looks like, what your fleet contains right now. It is not a language model trained from scratch; it is what makes a general model useful on WordPress. Its register of capabilities grows every day, and the counter on this page is that register.
No. The cost is shown before a draft runs; if the balance or a cap is short, the assistant stops and asks. Nothing is half-done on a site.
No. Your content is never sold and never used to train a model, by WebCrew or by a provider it uses.
Free for one site. No card.