• Who FusionSync is for
• Who FusionSync is not for
• The partnership model
• How FusionSync scales — infrastructure, not automations
• Product stack
• What this repository is (AIOS)
• AIOS + RecallSync — how we operate at scale
• Architecture
• Repository structure
• Getting started (operators & developers)
• Standard Operating Procedures (SOPs)
• Security & secrets
• Contact & credibility

FusionSync partners with agency owners who already have paying clients in one vertical and a daily tech gap — clients ask for software, CRM bridges, integrations, or AI workflows the agency cannot ship in-house.

Ideal partner profile:

Positioning in one line: FusionSync partners with agency owners who already serve a vertical, white-labels the product, wires the integrations, and runs white-glove implementation for each of their clients — so the agency can productize an offer it could not build alone.

FusionSync is not a freelancer marketplace, chatbot agency, or solo-founder prototype shop. We do not cold-pitch strangers with no client base.

FusionSync does not sell software to end clients directly. It partners with agency owners who already have paying clients in a vertical.

Typical partner engagement: $20k+ (scope depends on vertical, client count, and phase).

Division of work:

FusionSync stays the silent technical partner throughout. The agency owner stays the face of the business.

Current flagship vertical: dental patient acquisition — GoHighLevel, n8n, CRMBridge (HIPAA-compliant access to 27+ dental PMS systems), and WaCRM. Proof-of-model before replication into other verticals.

Proof on site: UK dental agency partner, creator monetization build, real estate underwriting, vertical SaaS fulfilment. Shubham Kashyap leads every partnership; team of 5 ships.

Most agencies scale tech delivery by stacking more automations — another n8n workflow, another Zap, another SaaS seat. That works for the first few clients. It breaks at ten.

FusionSync scales on owned infrastructure:

1. Automations are the wedge, not the ceiling. GoHighLevel + n8n prove demand fast (free POC, per-client rollout). They are the sales demo and first fulfilment layer — not the long-term product.
2. Production runs on self-hosted, open-source stack we control. RecallSync (CRM + AI agents + MCP), WaCRM (WhatsApp CRM + flows), Dograh (self-hosted voice agents), CRMBridge (dental PMS API layer), LiveKit, PostgreSQL, and Next.js — deployed under the partner's brand, not rented per-seat from closed vendors.
3. Repeatable fulfilment, not bespoke quotes every week. Productized tiers, fixed timelines, volume discounts on support — because the stack is the same; only vertical config changes.
4. AI-native operations. This repo (AIOS) drives agents, campaigns, and SOPs from Git via RecallSync MCP/CLI — so scaling delivery means scaling procedures and infrastructure, not headcount on manual prompt editing.

   Phase 1 — prove demand          Phase 2+ — own the stack
   ─────────────────────          ──────────────────────────
   GHL · n8n · integrations  ──▶  RecallSync · WaCRM · Dograh · CRMBridge
   (POC + first clients)          (white-label product + tenant isolation)
                                         ▲
                                         │
                              AIOS (this repo) — SOPs, context, agent ops via MCP

Why infrastructure beats automation-only scaling:

Delivery & integration layer: GoHighLevel · n8n · Make · Next.js · TypeScript · PostgreSQL · Prisma · Supabase · Stripe · OpenAI · Claude · LiveKit · Restack · Vercel

Roadmap: Google Reviews automation, content/social automation, local SEO tracking, Meta Ads integration.

This repo is the AI Operating System (AIOS) for FusionSync — the Git-driven control plane for running a business's AI agents, campaigns, and repeatable operations.

One repo = one business = one RecallSync sub-account.

Most teams operate AI agents through dashboards with no history, no review, and no single source of truth. AIOS treats business operations as code where it helps — while agent behavior stays in RecallSync, driven via MCP or the recallsync CLI.

RecallSync is the runtime. AIOS is the brain and playbook layer. Together they let FusionSync fulfil many clients without scaling manual ops linearly.

   define in RecallSync  ──▶  test via MCP/CLI  ──▶  run on channels
   (prompt / flow)          (test lead)            (live delivery)
        ▲
        └──  SOPs + context in Git guide how agents are built & operated

Core loop:

1. Define agent behavior in RecallSync (UI, MCP, or CLI).
2. Fetch the latest live state before editing (get-channel-agent).
3. Test in-place using the test lead harness; iterate until approved.
4. Activate when ready (update-channel-agent / UI).
5. Operate using SOPs and context files in this repo — repeatable across partners and verticals.

Local lead factory: The root Next.js CRM app (localhost:3010) stages and enriches agency partnership prospects before outreach. Enrichment reflects vertical, client count, tech gap, and integration targets — not generic founder MVP copy.

FusionSync Primary channel agents (live-chat, WhatsApp, Instagram, WP Call) carry the public partnership positioning; they are authored and tested live in RecallSync via MCP/CLI.

AIOS sits on a three-repo chain. This repo never talks to the database directly.

AIOS (this repo)  ──MCP──▶  recallsync-mcp  ──REST (Bearer api_key)──▶  recallsync-app  ──Prisma──▶  DB
   Cursor / CLI              thin tool layer         business routers              MySQL

See recallsync/ for the system/engineering layer, including add-mcp-operation.md.

Core concepts (RecallSync schema):

fusionsyncai/
├── aios/                   # rolling session context & intent-signal monitor
├── agents/                 # README only — agents live in RecallSync
├── context/                # business context agents read before acting
├── sops/                   # Standard Operating Procedures (repeatable checklists)
│   └── channel-agent/      # creation, prompting, tool-calls, testing, troubleshooting
├── playbooks/              # campaign definitions (objective, ICP, offer, sequence)
├── sprints/                # current build focus + decision log
├── recallsync/             # MCP integration docs & coverage map
├── scripts/                # recallsync CLI wrapper, reconcile-flow, precommit-sanity
├── telegram-bridge/        # optional Telegram → cursor-agent bridge
├── app/                    # local AIOS CRM (lead factory, port 3010)
└── .env.example            # secret registry — names + purpose only (no values)

• Cursor — this repo is designed to be operated by its agent
• RecallSync sub-account and API key
• recallsync-mcp server running (or hosted endpoint)

cp .env.example .env.local   # .env.local is gitignored

Set RECALL_API_KEY and RECALLSYNC_BASE_URL (hosted RecallSync domain, or http://localhost:3000/api/rest for local recallsync-app).

Template: .cursor/.mcp.example.json. Real config: .cursor/mcp.json (gitignored).

Typical first tasks:

• "List the primary agents." → confirms MCP/CLI connection
• "Create a WhatsApp channel agent." → follows sops/channel-agent/creation.md
• "Update the live-chat agent flow." → fetch live, edit, push draft via MCP/CLI

CLI (preferred for agent sessions):

node scripts/recallsync-cli.mjs --json primary-agent list
node scripts/recallsync-cli.mjs --json channel-agent get --id <id>

Run AIOS from your phone — whitelisted Telegram DM spawns cursor-agent with RecallSync MCP tools. See telegram-bridge/README.md.

An SOP is a repeatable checklist a human or AI agent can follow literally. If something is done more than once, it becomes an SOP.

Start a new SOP from sops/_template.md.

• Gitignored: .env* (except .env.example), .cursor/mcp.json, *.pem
• Placeholders in committed files: ${AIOS_N8N_WEBHOOK_TOKEN} — resolved and encrypted at push time
• API guard: RecallSync rejects headers still containing ${...} placeholders
• Pre-commit: bash scripts/precommit-sanity.sh must print RESULT: ✓ SAFE before commit

Full procedure: sops/git-commit.md.

For agency owners with paying clients in a vertical: message on WhatsApp with your vertical, client count, and the tech gap you cannot fill today. We map phased partnership fit before you commit.