Website Agent
An AI system with direct, live access to your web server. Instead of emailing a developer and waiting, you type what you need in plain language—the agent reads the files, makes the changes, and confirms the result.
How it works
Your website lives on a managed server. The agent has authenticated file access to that server.
You communicate through Swift Chat—a private, dedicated chat interface configured for your business.
You type a request—“update the homepage hero text,” “add a new service page,” “fix the contact form”—and the agent executes it directly.
Every change is visible and discussable in your Swift Chat room. Chat history is retained—you have a record of every change and why.
The result: a responsive, always-available technical partner that operates without a ticket queue, a dev retainer, or a delay.
We engineer first
Before the agent runs freely, our team defines structure: templates, navigation, SEO baselines, forms, and deployment rules. The agent then generates and revises pages within those boundaries—drafting copy, suggesting layouts, and opening merge-style reviews when you want human approval.
The agent reads your approved sitemap, style constraints, and business facts you provide. It doesn't “invent” a second website—it extends the one we designed together, keeping listings, hours, bios, and offers aligned with what you've validated.
Included capabilities
- ✓ File-level access to your hosted website
- ✓ Plain-language change requests—edit, rewrite, add, remove
- ✓ Page creation, content editing, structural updates
- ✓ Contact form submissions routed to your Swift Chat channel
- ✓ Error triage and basic troubleshooting
- ✓ SEO hygiene: metadata, redirects, sitemaps
Infrastructure included
- ● Managed VPS or cloud server hosting
- ● Domain DNS configuration
- ● SSL certificate management
- ● Swift Chat deployment (your dedicated room)
- ● Agent endpoint configuration
- ● Base utility library (built once, no token cost)
Understanding inference costs
Every instruction you give the agent requires the AI model to read context, reason through the task, and generate a response. This is measured in tokens (roughly three quarters of a word). Inference costs are passed through at our sourced rate—no markup.
Costs tokens
- ✓ Each instruction you give the agent
- ✓ Agent reading files to understand current state
- ✓ Agent reasoning and writing a response or edit
No cost
- ✗ Utility functions (built once)
- ✗ Your server responding to visitors
- ✗ Swift Chat messages
- ✗ Onboarding and context updates from our team
Real-world costs
A full active development day—multiple sessions, numerous edits to a live website—typically costs $5–$8 in raw inference. Typical operational days run $0.50–$2.00. Expect the first 1–3 months to represent peak inference spend, with costs normalizing as the agent's context matures.
Your AI Account Manager
A real person is assigned to your account. They handle context engineering—building and maintaining the instruction set, memory, and rules that make your agent accurate for your specific business. They also manage model selection, cost oversight, and strategic advisory. This is the human in the loop.
Context engineering is the core of what we do. Well-engineered context produces an agent that feels like it understands your business—and that intelligence compounds over time. After 3 months, you can nominate a team member for our AI Account Manager Certification to bring this capability in-house.
Investment
One-time setup
$550–$1,100 avg
Covers typical server provisioning, Swift Chat, agent utilities, and initial context engineering. Large projects—big migrations, multi-site, or heavy custom build-out—can be much more; we scope and quote to the work.
- Server provisioning & environment setup
- Swift Chat deployment & configuration
- Agent utility library build
- Context engineering—initial onboarding
- Website migration or initial site build quoted separately if needed
Base
Managed hosting
from $50/mo
VPS / managed hosting sized for typical small-business traffic. Monthly cost rises with backup depth, retention, compliance needs, and other infrastructure objectives—we quote to what you actually need.
- Server hosting & maintenance
- SSL, DNS, and backups
- Security updates & uptime monitoring
- Swift Chat room for support requests
For a site launched and reliably hosted—AI optional.
Core
Basic inference costs
from $30/mo
Monthly floor for Website Agent access and baseline inference. Usage beyond that is passed through at cost (see inference above).
- Plain-language site changes via the agent
- Inference beyond the floor billed as you use it
- Pairs with managed hosting; strategy tier adds full context + human oversight
Entry point for ongoing agent work on your hosted site.
Add-on
AI Agent & strategy
$950–$2,150/mo
- Builds on basic inference (Website Agent + floor), with sustained depth
- Ongoing context management & instruction tuning
- AI Account Manager (human in the loop)
- AI strategy & advisory
- Inference pass-through on top (typical add-on still in the ~$15–$40/mo range for many sites)
For teams that want an active AI partner and a named human managing it.
Basic inference costs start at $30/month. Managed VPS hosting starts at $50/month for standard small-business traffic; pricing increases with backup requirements, retention, and your broader business objectives.
Add-on
Custom software & integrations
Scoped
- Custom API endpoints and integrations
- Booking, scheduling, or lead qualification agents
- CRM, payment, or inventory connections
- Multi-agent architectures
- Internal tools for your team
Built as a discrete project or folded into the monthly managed service.
Every engagement starts with managed hosting. Add Basic inference costs from $30/mo when you want the Website Agent on your stack; layer in AI Agent & strategy for full context engineering and an AI Account Manager. Custom integrations are scoped separately. You'll get a specific monthly estimate after discovery—site size, change volume, traffic, and backup needs all factor in.
Ready to scope your Website Agent?
Leave a message