Local quick answers
Approved prompts, saved replies, utility answers, and simple math can resolve common questions before any provider call is made.
Custom website AI chatbots
Custom AI chatbots for serious websites.
Canopy builds practical AI chat agents for buyer-facing websites: fast guided answers, server-side model calls, bilingual handoff, quote routing, and Canopy AI platform planning for Ontario and Canadian teams.
- Surface Website agent
- Runtime Server-side AI bridge
- Handoff Quote, payment, or human follow-up
Canopy AI platform plan
A funnel that can become the first production plan.
The page sells the visible chat box, but the implementation plan covers the knowledge source, server endpoint, handoff contract, usage controls, and what the owner needs to review after launch.
Server-side model calls
The browser sends only the prompt, locale, context, and honeypot field. Provider credentials and rate rules stay behind the PHP bridge.
Bilingual buyer handoff
English and French page copy, prompts, chat labels, and quote routes keep the public funnel usable for Canadian buyers.
Quote and payment routing
Serious conversations can become quote intake now, and payment/token lanes can stay staged until owner review approves them.
Ontario/Canada readiness
The build is framed around practical deployment controls, topic boundaries, fallback replies, rate limits, and maintainable handoff notes.
Buyer journey
Not just a chatbot. A web sales assistant with boundaries.
The first version should answer common questions, gather useful project context, route serious buyers, and avoid burning AI calls on prompts the site can answer instantly.
Answers before AI spend
Quick prompts, saved replies, basic utility answers, and simple math can resolve common interactions locally before the model is needed.
Server-side model bridge
Provider keys, rate limits, payment rules, and fallback replies stay server-side. Buyer messages can still be processed by the configured model provider.
Sales and payment handoff
The agent can move visitors into quote intake, a hosted payment flow, or a scoped client assistant when the conversation is ready.
Canopy website agent
Website AI agent
Usage model
Transparent model-cost-based pricing.
Public pages should not hardcode token prices because provider rates change. Canopy can scope usage from the current DeepSeek Pro base cost, confirm currency and provider reference at quote time, and show the multiplier plainly.
Standard website agent
DeepSeek Pro base cost x 1.25 for normal buyer conversations, plus the implementation and support scope agreed for the site.
Advanced assistant tier
DeepSeek Pro base cost x 1.5 for heavier workflows, richer context, more evaluation, or higher-touch operations.
No public key exposure
Usage accounting, provider credentials, checkout rules, and abuse controls stay in the server layer, not the browser.
AI chatbot quote path
Build the chatbot brief before opening the quote form.
Pick the business shape, agent job, website surface, handoff path, and usage plan. The quote form receives both the normal product-builder fields and the AI-specific context.
Canada-ready implementation
Built for practical Ontario and Canadian deployment.
- English and French website copy, prompts, suggestions, and handoff paths.
- Server-side provider calls so API keys and payment rules are not exposed in the browser.
- Local answers for low-value prompts such as test, ping, and arithmetic.
- Quote intake and payment-path planning for buyer conversations that become real work.
- Topic boundaries, fallback responses, rate limits, and reviewable public knowledge briefs.
- Deployment readiness means practical controls and handoff, not a legal compliance guarantee.
AI agent references
Related integration and chat surfaces
Setup path AI integration readiness checklist A consultant-led path from business use case to safe public or internal deployment.Chat Canopy AI chat A constrained chat surface for Canopy-related questions, gated behind provider environment config.Writing AI workflow articles Reference notes on model APIs, agent setup, harnesses, and local-first developer workflows.
Next step
Add a website agent where buyers already land.
Start with the buyer questions, approved answers, handoff rules, traffic expectations, and the places where human follow-up or payment should take over.