WordPress · Leaf Concierge
Leaf Concierge
Answers visitor questions from your own published pages, names the pages it used, and declines when your site has no answer.
How it answers
1. It looks through your site
Your published pages are cut into passages and kept current as you edit. Drafts and password-protected pages are never read.
2. It checks it has an answer
If your pages do not cover the question, it declines. No model is called and nothing is invented.
3. It cites what it used
The model must number its sources. Only pages the reply actually cites are shown beside the answer.
What you get
Content gaps
Every question it could not answer, counted and ranked. That list is your next set of pages to write.
Approved answers
Write exact wording for prices, policies and hours. Served instantly, with no model call and no drift.
Knows the page
Visitors can ask “summarise this page”. It prefers the page they are reading.
Team requests
When it cannot answer, visitors send the question to you, with consent, by email and in your dashboard.
Four providers, one failover
OpenAI, Claude, Google Gemini and OpenRouter. If one fails, the next answers, inside a fixed time limit.
Private by design
No IP address, name or account is stored with a question. Records expire on your schedule. Keys are encrypted, or kept in wp-config.php.
Light on your site
One deferred script and one small stylesheet. System fonts. No third-party request until a visitor asks.
You see what it did
Test any question in your dashboard: matched pages, match scores, whether a model was called, and why it declined.
Pricing
Every plan includes every feature. Plans differ only in how many sites the licence covers. A licence includes updates and support; the assistant keeps working if it lapses. You bring your own AI provider account and pay the provider directly.
- All features in every plan: citations, approved answers, content gaps, team requests, page awareness, four providers with failover.
- Updates and support while your licence is active.
- Prices include GST. Choose a billing period above; to change period later, buy the other one when your current key is due.
Questions
| Will it make things up? | In the default mode it cannot answer without matching passages, and the model is told to reply “no answer” if they do not contain one. A model can still misread a passage, which is why the pages it used appear beside every answer. |
| What does it cost to run? | Only the model calls you make, billed by your provider. Greetings, approved answers and declined questions make no call. Your dashboard shows token counts for the last 30 days. |
| Does it work with page builders? | It indexes the content WordPress stores for each page. Text built dynamically at display time is not stored there and is not indexed. |
| What does it send to the AI provider? | The visitor’s question, up to six earlier messages from the same chat, and matching passages from your own pages. No IP address and no account details. |
| Requirements | WordPress 6.0 or later, PHP 7.4 or later, and an API key from one supported provider. |