Answers with receipts
Every reply lists the pages it was built from, as links the visitor can open. If nothing in your content matches, Askora says so instead of guessing.
★ Launching on CodeCanyon
Askora reads the pages you have already written, answers visitors from them, and shows the links it used. It never invents a price. When someone is ready to buy it asks for their details, and every question your site could not answer lands in a report you can act on. Your API key, your database, no monthly fee per conversation.
Overview
Most AI chat plugins are a chat box with a general-purpose model behind it. Ask one a question about your prices and it will happily make a number up, because it has never read your pricing page. That is the failure that costs you a customer, and it is the one Askora is built around.
Askora splits your published pages into passages, embeds them once, and keeps them in your own database. A question is matched against those passages first; only the ones that actually match are sent to the model, with an instruction to answer from them or say it does not know. Every answer carries the links it came from, so a visitor — and you — can check it. Prices, refund terms and opening hours can be pinned word for word, and those answers never reach a model at all.
Screenshots
Pricing
One-time
The whole plugin on one site, with updates. No monthly fee to us and no per-answer fee — you bring your own API key and pay your provider directly.
Quoted
Askora is what we needed often enough to package. When a project needs something it does not do, we build that instead.
Askora is a commercial plugin and is going on CodeCanyon as a one-time purchase — no subscription to us, and no charge per conversation. What the answers cost is what your own provider charges: on Google Gemini's free tier, a small site runs at nothing. Early access is open now; tell us about your site and we will send you the build and the price before it lists.
Changelog
Reasoning models spend the answer allowance on their own thinking, which returned an empty reply. The limit is now topped up automatically, and any bookkeeping a model repeats under its answer is stripped before the visitor sees it.
Every Askora screen now reads as one product: the numbers you came for as tiles at the top, plain status labels, roomier tables, and empty screens that say what will appear there and why.
A quieter header, softer panel, animated typing dots, sources as chips, a dark theme that can follow the visitor's own device — and a Design screen that previews the real widget as you change it.
Turn an unanswered question into a drafted answer in one click, then keep it as a pinned answer or save it as a draft post for you to edit and publish.
Pin an exact answer for the questions that must never be re-worded, matched by meaning as well as by wording. Set a monthly token budget and the chat stops rather than overspending.
Point Askora at OpenRouter, Groq, or a model running on your own machine, and use that service's key.
Purchase-code activation with a grace period, and update checks gated on the licence.
Questions
Questions about a custom plugin built for your site are on the plugin development page.
Your own published content, and nothing else. Askora splits each page into passages, embeds them once, and stores them in your database. When a visitor asks something, the closest passages are found first and only those are sent to the model, with an instruction to answer from them or admit it does not know.
That is why every answer carries links: they are the pages it actually used. If nothing in your content matches the question, Askora says so instead of filling the gap with something plausible.
Askora itself is a one-time purchase. The answers cost whatever your own provider charges, billed to your own account — there is no fee to us per conversation and no data passing through us. Google Gemini has a free tier that covers a small site outright, and indexing is a one-off cost per page that only repeats when you edit that page.
There is a monthly token budget in Settings: set it and the chat stops when it is reached, so a surprise bill is not possible.
Google Gemini, OpenAI and Anthropic are built in. There is also a custom endpoint field for anything that speaks the OpenAI protocol — OpenRouter, Groq, Together, or a model running on your own server — with that service's own key.
Indexing needs a provider that produces embeddings, so Gemini or OpenAI. Anthropic can write answers but cannot build the index.
It should not, and there is a second line of defence for the answers where it matters most. Pin the exact wording for a price, a refund term or your opening hours, and Askora returns that text as written, before any model is involved — which is also instant and free.
A pinned answer matches questions that mean the same thing, not just the exact sentence you typed.
The widget is vanilla JavaScript and CSS, under 25KB, and renders inside a shadow root so it cannot inherit or leak styles. There is no framework, no iframe and no third-party script. Nothing is loaded until the page has finished its own work.
Indexing runs in batches sized to what your host actually allows, and backs off on its own when a provider rate-limits it — it is built for shared hosting, not a dedicated server.
Conversations and leads are stored in your own WordPress database. The only thing that leaves your server is the question and the matching passages, sent to the AI provider you chose, on your own account. API keys are stored encrypted using your site salts. The conversation log can be turned off entirely, and there is a per-IP hourly rate limit as standard.
WordPress 6.4 or newer and PHP 8.0 or newer, on ordinary shared hosting. No Node, no Composer, no vector database, no background workers and no shell access. An API key from your chosen provider is the only other requirement.