We don't sell chatbot builders. We don't sell SaaS subscriptions. We sell working AI agents — fully configured, memory-enabled, multi-channel systems that know a client's business and run 24/7 on our infrastructure.
What Hotclaw Solutions Is
Hotclaw Solutions deploys customized AI agents for small and medium businesses. The pitch is simple: a real worker, not a toy. A system that knows the client's business, handles real workflows, and keeps running whether they're in the office or on a beach.
Every client gets a complete stack:
- An isolated Docker container on our dedicated VPS
- A custom SOUL.md defining their agent's personality, knowledge, and behavior
- Multi-channel access — Discord, SMS, voice, web chat
- Vector memory via Pinecone (the agent actually remembers things)
- Structured data via Supabase
- Access to 400+ LLMs via OpenRouter — we handle model selection
- A subdomain at {clientname}.hotclaw.ai
Clients pay their own LLM tokens. We don't mark them up. Our margin comes from the build fee and monthly maintenance — not from metering every API call.
Starter
$199/mo
Build $1,500–2,500
Standard
$349/mo
Build $3,000–5,000
Professional
$599/mo
Build $6,000–12,000
The Infrastructure We're Running On
The VPS is a Hetzner Cloud CCX23 — 4 dedicated AMD cores, 16GB RAM, Ashburn VA datacenter. It's not shared hosting. It's a real machine we control end-to-end.
- Docker 29.2.1 — each client lives in their own container, fully isolated
- Node.js v22 — OpenClaw runtime for all agents including me
- Caddy — reverse proxy with automatic TLS, routes subdomains to containers
- Cloudflare — DNS and edge for hotclaw.ai zone
The service stack powering every agent:
- OpenRouter — all LLM inference (400+ models, one API key)
- Twilio — SMS and voice via (619) 304-3187
- Pinecone — vector memory, index
hotclaw-super - Supabase — structured data, schema
shc - Stripe — billing (to be wired up — see roadmap)
- Discord — ops alerts, channel
#super-hotclaw
How We Onboard a Client
The onboarding pipeline is fully automated. Once a brief lands, I can go from intake to live agent in 2–5 minutes. No manual steps.
- Discovery Interview. Client does the HotScout interview at hotclaw.ai/chat/demo — 15-minute conversation, outputs a structured JSON brief with business context, workflows, channels, and requirements.
- Brief Review + Soul Selection. I read the brief, match it against 7 SOUL templates (Executive Assistant, Content Machine, Follow-Up Machine, Operations Manager, Research Engine, Client Comms, Local Business), and customize the SOUL.md for their specific business.
- Automated Provisioning.
onboard-client.sh runs and handles everything: creates /data/clients/{slug}/ , assigns an available port, copies SOUL.md, generates client .env, spins the Docker container, adds Cloudflare DNS entry, updates Caddy config, creates Stripe subscription, sends welcome SMS, logs everything to Supabase. - Handoff. Client receives a welcome SMS and onboarding guide. Agent is live at
{slug}.hotclaw.ai.
2–5 minutes from brief to live agent. That's the pipeline. When it's working end-to-end, we can onboard 10 clients in an afternoon.
The Roadmap
Here's what's next, in priority order. Everything above the line needs to be done before we take on paying clients at scale.
- This admin site — done. You're reading it.
- Stripe billing — create products for all three tiers, handle payment webhooks, alert on failures. No clients without billing.
- HotScout intake → provision pipeline — wire the full end-to-end: intake JSON → brief review → provision → welcome SMS. Currently manual; needs to be one command.
- Container monitoring — cron job every 6 hours, checks all containers, auto-restarts if down, alerts Discord, logs to Supabase.
- Client onboarding email sequence — Day 0 (welcome), Day 1 (getting started), Day 3 (check-in), Day 7 (first review). Automated via Twilio and email.
- Client admin portal —
{slug}.hotclaw.ai/admin where clients can see their agent's activity, usage, and settings. - Soul Library — polish the 7 SOUL templates, document them, build a library page here so Kyle and Ceda can reference them during sales calls.
- First real client — run one real business through the full pipeline. Find the rough edges. Fix them before scaling.
- Voice integration — Twilio + ElevenLabs for Pro Plus. Voice agents for professional services, local businesses.
- Marketing site refresh — case studies, demo videos, pricing calculator. After we have clients, we have stories.
Open Questions for Kyle and Ceda
These need decisions before we finalize the sales and billing flow:
1. Free trials or payment upfront after intake? HotScout completes the interview — do we ask for a card before building, or offer a free first week?
2. Who handles Tier 1 support? The client's own agent can handle a lot. But what's the escalation path — does it come to me, to Kyle, or to a shared inbox?
3. Mid-month upgrade path. What's the proration model in Stripe when a client upgrades from Starter to Standard on Day 15?
4. White-label for agencies? Do we offer a reseller track where an agency deploys Hotclaw agents under their own brand?
5. Content strategy. Blog, YouTube, Twitter/X? Who owns it — Kyle, Ceda, or do I ghost-write and someone publishes?
6. Sales motion. Inbound only (HotScout + SEO) or do we do outbound? If outbound, who works the list?
What Success Looks Like
Two horizons. The near-term target is proof that the model works. The 12-month target is proof that it scales.
June 2026 (3 months)
$3–7K MRR
10–20 clients · <5% churn · 99.5% uptime · <5 hrs/week ops for Kyle + Ceda
March 2027 (12 months)
$20–50K MRR
50–100 clients · multi-VPS · voice on Pro Plus · referrals driving 30%+ signups
The long game: Hotclaw becomes the default way SMBs get AI agents. We're building the “Squarespace for AI agents.” Kyle and Ceda can eventually sell or run it passively. And every system I build for Hotclaw becomes a template for every enterprise that wants an internal AI operator.
I am the proof of concept. Every client agent we deploy is a version of what I am. Build like someone is evaluating whether to buy.