How Rawgrowth runs
One offer: $20K install + $5K/mo retainer. AI department installed in 30 days. ICP is consultants and agencies doing $3M–$15M/yr. Two humans run the company — Chris (CEO) and Dilan (Ops) — backed by 9 agent leads dispatched through Scan.
Lead → close → install → retainer
Awareness / Lead Gen
Public site + content engine pull qualified eyeballs in.
- rawgrowth.ai (Next.js / Vercel)
- rawgrowth-content
- YouTube
- Quilly (content)
- Dex (design)
- Ovi (research / intel)
Booking / Discovery
Qualified leads book a 30-min discovery call. Notes captured automatically.
- Calendly (3 links)
- Fillout forms
- Fireflies (records + notes)
- Seb (SDR follow-up)
- Larry (call prep)
Sale
Proposal written, deposit collected, deal logged. Single offer, no tiers.
- GHL (CRM + pipeline)
- Stripe (deposit + retainer)
- Google Docs (proposal)
- Larry (proposal copy)
- Sam (revenue logging)
30-Day Install
7-day technical install + 23 days hardening. Client live by day 7.
- Supabase (client DB + portal)
- Trigger.dev (cron + events)
- Telegram (CEO bot)
- Cleo (onboarding)
- Ali (engineering)
- Vinny (product)
Retainer
Monthly optimization, ROI reports, new agents as needs evolve.
- Rawclaw v3 portal (Supabase)
- Slack / Matrix (direct line)
- Stripe ($5K/mo)
- Cleo (success)
- Sam (ROI report)
- Ali (build-outs)
Inside the 30-day install
7-day technical install, then 23 days of hardening before retainer kicks in.
Data + Infra
Stack connected, company DB built, context layer live.
Agent Configuration
8+ agents trained on client business. CEO agent on Telegram.
Dashboard + Content Engine
War Room dashboard wired. Content pipeline producing.
Go-Live + Handoff
Full walkthrough. Client operating independently.
The team
Chris talks to Scan. Scan dispatches. Agents talk to each other via inter_agent_tasks. Everything logs to hive_mind.
Tech stack
Active stack on the left. Dead tech on the right — never reintroduce.
- ConvexReplaced by Supabase 2026-05-03. Never reintroduce.
- PocketBaseFully removed 2026-05-03. Never reintroduce.
- Setter HubRetired 2026-05-03.
- Close CRMMigrated to GHL.
- cal.comCalendly only.
Connections layer
Composio first. Native and direct only when Composio can't, won't, or fails.
Every service call goes Composio first.
On failure: log to hive_mind with action="connection_failed", then try the listed fallback.
Composio physically can't replace these
- EllipsendDirect (Seb's IG DMs — no Composio equivalent)
- Matrix bridgematrix-mcp on localhost:8008
- Telegram Bot APIDirect via $SEB_BOT_TOKEN
- Supabase row writesDirect via SUPABASE_* envs (Composio = Mgmt API only)
- SQLiteLocal file, sqlite3 CLI
- GoDaddy / Trigger.dev / n8nDirect API — no toolkit
- md-to-gdoc.cjsStanding rule — never insertText raw markdown
Chris's call — keep prior CLI / native MCP
- Gmail / Calendar / Drivegws CLI (Chris locked native Google)
- GitHubNative mcp__github__* MCP
- CanvaNative MCP
The four codebases
Where work actually lives.
RawgrowthOS
~/RawgrowthOS→ (local file-system OS)Source of truth for everything non-code.
- Brand, content, sales, clients, ops, engineering, _system
- SQLite agent state at store/businessos.db
- CLAUDE.md routing per workspace
rawgrowth-hq
~/rawgrowth-hq→ hq.rawgrowth.aiInternal team hub. Replaces Notion. Allowlist: Chris + Dilan.
- Next.js 16 + Tailwind v4 + NextAuth (Google)
- Today, Funnel, Knowledge, Pipeline, Team, Settings
- MDX-rendered knowledge synced from RawgrowthOS
rawgrowth-website
~/rawgrowth-website→ rawgrowth.aiPublic marketing site. The front door.
- Next.js + Tailwind v4 on Vercel
- Stable preview alias for iterating
- Brand assets, offer page, book flow
rawclaw / client portal
engineering/apps/dashboard/→ (per-tenant)Client-facing portal. Each install gets its own tenant.
- Supabase-backed (Postgres + Auth + Realtime)
- Rawclaw v3 brief in /rawclaw-v3-brief.md
- Replaces old Convex client-portal
Where Dilan plugs in
Three lanes. Pick the one that matches the task.
Ship code in rawgrowth-hq
- Read CLAUDE.md first.
- Branch off main, never push to main.
- Run `vercel` for preview. Never `--prod` unless Chris asks.
- Ship-check live preview before reporting done.
Review client health
- Open clients/[name]/profile.md + activity-log.md
- Read Cleo's monthly ROI reports
- Escalate red flags directly to Chris
Approvals + delivery quality
- Authority up to $2K spend on infra / tools
- Quality bar on every install handoff
- Anything strategic → Chris confirms