What the Client Gets
A structured 10-page report covering: competitor analysis, Instagram/DM audit, revenue leak breakdown, top 5 action items, and an AI automation roadmap. Delivered as a PDF within 48 hours of intake form submission.
The target buyer is a beauty pro, salon owner, or solo operator who wants to understand where AI can help their business before spending money on tools. The audit IS the pre-sale diagnostic — it builds trust before any larger engagement.
The Stack
| Layer | Tool | Role |
|---|---|---|
| Payment | Stripe | $297 payment link — client pays upfront |
| Intake | Formspree + Netlify | Client fills business details form |
| Research Engine | DeerFlow 2.0 | Multi-agent research framework (4 Docker containers) |
| Research Model | Claude Sonnet 4.5 | Backing model for DeerFlow agents |
| Custom Skill | business-audit skill | DeerFlow skill that structures the audit research |
| Orchestration | Claude Code | DYOE Way runs the audit, reviews the output, edits the report |
| Delivery | Email (manual) | DYOE Way sends the final report to the client |
How DeerFlow Works
DeerFlow is an open-source multi-agent research framework. We self-host it on Docker (4 containers: nginx, frontend, gateway, langgraph) locally and on a DigitalOcean droplet. The framework coordinates multiple Claude agents to run parallel research tasks and synthesize findings.
We wrote a custom business-audit skill — a structured prompt that tells DeerFlow what to research, in what order, and how to format the output. The skill takes the client's intake data as input and produces a structured research package as output.
The Audit Pipeline
- Client pays via Stripe → DYOE Way gets Stripe notification
- Client fills intake form at dyoeway-audit.netlify.app → DYOE Way gets Formspree email
- DYOE Way opens DeerFlow (browser or MCP), runs the business-audit skill with client data
- DeerFlow agents (running Sonnet) pull public signals: competitor websites, Instagram profiles, Google Business, review sites, industry benchmarks
- DeerFlow synthesizes findings into the 10-page report structure
- DYOE Way reviews the entire report, edits for accuracy and tone, adds specific action items
- DYOE Way delivers the report to the client via email within 48 hours
Why Self-Hosted
We run DeerFlow on our own infrastructure for three reasons:
- Cost control — no per-query SaaS fees; the only variable cost is the Claude API usage
- Privacy — client data never touches a third-party research platform; it stays on our controlled infrastructure
- Customization — we wrote the business-audit skill ourselves; the research patterns are tuned for beauty/salon/service businesses, not generic
MCP Integration
DeerFlow is also registered as an MCP (Model Context Protocol) server in our Claude Code environment. This means DYOE Way can trigger a business audit from within Claude Code without opening a browser — useful for batch processing or when iterating on the skill.
The MCP server exposes five tools: deerflow_health, deerflow_list_skills, deerflow_research, deerflow_get_result, and deerflow_run_skill.
Economics
At $297 per audit, the math only works if delivery cost is low and predictable. With DeerFlow on our own server and Claude API costs managed by model routing (Haiku for classification, Sonnet for research), the per-audit API cost is a small fraction of revenue. The highest cost is founder time reviewing and editing — which is the point. The human review IS the product quality.
What's Next
We're building a Telegram integration so audits can be kicked off from anywhere via the /research command in the DYOE Agent Stack bot. The research runs async on the DigitalOcean droplet and results come back to Telegram when complete. This cuts the turnaround from "sit at the computer" to "start it from anywhere."