AIECOS Social CRM
Open-source template to sync Pancake (Zalo OA / Facebook Messenger / Instagram) data into your own Supabase, with a built-in admin UI, MCP server for AI agents, and B2B partner classification.
π Live demo Β· π Setup guide Β· π€ MCP usage
π¬ Demo
https://github.com/aiecosvietnam/aiecos-social-crm/releases/download/v1.0.0/demo.mp4
11-second walkthrough β Dashboard Β· Inbox Β· Pipeline Β· Partner 360 Β· Performance.
Inline player auto-renders on GitHub.com. Fallback: direct download.
What it does:
- π₯ Chrome extension reads Pancake DOM β POST to your sync receiver
- πΎ Sync receiver writes to Supabase schema (
pages,customers,conversations,messages) - π₯ Standalone admin UI (single HTML file) reads from Supabase REST API
- π€ MCP server lets Claude / any MCP client query data directly via natural language
- π 5-stage partner classification: Active / Sleeping / At-Risk / Dormant / Churned
- π Export CSV + printable HTML reports
- π¬ Built-in Inbox (3-pane), Triage alerts, Pipeline kanban
- π Demo mode β try the UI instantly without setting up Supabase
Stack: Node.js + Express, Supabase (Postgres + PostgREST), Chrome Manifest V3, MCP SDK.
Architecture
βββββββββββββββββββ POST ββββββββββββββββββββ upsert ββββββββββββββββ
β Pancake DOM β ββββββββββΆ β Sync Receiver β βββββββββΆ β Supabase β
β (Chrome ext) β β (Node + Express) β β aiecos_socialβ
βββββββββββββββββββ ββββββββββββββββββββ ββββββββ¬ββββββββ
β PostgREST
ββββββββββββββββββββββββββββββββ
βΌ
ββββββββββββββββ ββββββββββββββββ
β Admin UI β β MCP Server β
β (HTML/JS) β β (for Claude) β
ββββββββββββββββ ββββββββββββββββ
Screenshots
Admin UI β Dashboard (live data from your own Supabase):
- π KPI cards: Total partners / Active / At-Risk / Dormant+Churned / Total messages
- π 14-day activity trend (Customer vs Agent split)
- π¬ Recent messages stream
Pipeline Kanban β 5 stages with auto-classification:
βββββββββββ ββββββββββββ βββββββββββ βββββββββββ βββββββββββ
β ACTIVE β β SLEEPING β β AT-RISK β β DORMANT β β CHURNED β
β β€ 3d β β 3-7d β β 7-30d β β 30-90d β β >90d β
βββββββββββ ββββββββββββ βββββββββββ βββββββββββ βββββββββββ
Triage β Auto-alerts for partners going silent Β· Partner 360 β Full table Β· Performance β Customer/Agent ratio + top partners
β Try the live demo (connect to your Supabase via Settings)
Folder structure
aiecos-social-crm/
βββ README.md
βββ SETUP.md β Start here
βββ LICENSE β MIT
βββ docker-compose.yml β One-command dev stack
βββ .env.example β Root env (for docker compose)
βββ chrome-extension/ β AIECOS Pancake Connector (Manifest V3)
βββ sync-receiver/ β Express server + Dockerfile + schema.sql
βββ mcp-server/ β MCP server for Claude
βββ admin-ui/ β Single-file HTML dashboard
βββ examples/ β Curl scripts, seed data, MCP prompts
βββ docs/
β βββ DEPLOY.md β Production deployment guide
β βββ MCP_USAGE.md β MCP tool reference
β βββ ARCHITECTURE.md β Internals
βββ .github/workflows/
βββ ci.yml β Syntax + secret scan + docker build
Quick start (1 minute β Docker)
# Boot the entire stack: Postgres + PostgREST + sync receiver + admin UI
cp .env.example .env
docker compose up -d
# Inject demo data (5 partners across all stages)
bash examples/seed-demo-data.sh
# Open admin UI
open http://localhost:8080
# β Settings β Supabase URL: http://localhost:3000 β Schema: aiecos_social β Save
Done. You can now see Active / Sleeping / At-Risk / Dormant / Churned partners in the kanban.
Quick start (manual, step-by-step)
# 1. Set up Supabase (cloud or self-host) + run schema.sql
psql -f sync-receiver/schema.sql
# 2. Start sync receiver
cd sync-receiver
cp .env.example .env # edit with your Supabase credentials
npm install
npm start
# 3. Open admin UI in browser
open admin-ui/index.html
# β Click Settings β paste your Supabase URL + anon key
# 4. Install Chrome extension
# chrome://extensions/ β Load unpacked β select chrome-extension/
# 5. (Optional) Wire MCP server to Claude
cd mcp-server && npm install
# Add to ~/.claude.json mcpServers, then restart Claude Code
Full step-by-step instructions: SETUP.md
Why this template?
Most CRMs lock you into their data silo. This template gives you:
| Feature | Benefit |
|---|---|
| Own your data | Self-host Supabase, full Postgres access |
| AI-ready | MCP server exposes data to Claude / any LLM |
| Multi-channel | Facebook + Zalo OA + Instagram via single Pancake account |
| Zero vendor lock | All code MIT, no proprietary deps |
| B2B-aware | Partner classification (Active β Churned) out of the box |
| Audit trail | Every message logged with timestamp + sender_type |
Use cases
- B2B distributors: Track shop partners, alert when they go silent
- D2C brands: Multi-channel inbox aggregation
- Agencies: White-label social CRM for clients
- AI assistants: Feed conversation history to your AI agents
Roadmap
- [ ] Add HubSpot / Salesforce sync
- [ ] Webhook out (Slack/Telegram alerts)
- [ ] Tag/segment management UI
- [ ] LLM-powered reply suggestions
β Support development
NαΊΏu template nΓ y giΓΊp Γch cho bαΊ‘n, α»§ng hα» mα»t ly cΓ phΓͺ Δα» mΓ¬nh tiαΊΏp tα»₯c build thΓͺm nhiα»u dα»± Γ‘n mΓ£ nguα»n mα» thα»±c tαΊΏ cho cα»ng Δα»ng dev Viα»t Nam.
If this template saved you time, consider supporting future open-source work. 100% goes back into building more real-world templates (CRM, AI agents, integrations like KiotViet/MISA/HubSpot/Shopify, Vietnamese + English docs).
π»π³ MoMo Β· VietQR Β· Napas247
NGUYEN TAN HOANG
PSG2614514200000011 Β· NH: MoMo
Scan with MoMo / any VietQR-compatible app (Vietcombank, Techcombank, MB, TPBank, β¦)
π Other ways to support
| Action | Why it helps |
|---|---|
| β Star this repo | Boosts visibility β more devs find it β more contributors |
| π Open issues / PRs | Real-world feedback shapes the roadmap |
| π’ Share with your network | Especially if you work on Vietnam social-commerce |
| πΌ Hire AIECOS for custom builds | aiecos.ai β AI Sale Agent / AI Customer Care / custom CRM |
| π¦ Tag us when shipping | Twitter / LinkedIn β we'll amplify |
CαΊ£m Ζ‘n π β every star, comment, donation keeps this momentum going.
Credits
Built by AIECOS β open-source AI infrastructure for Vietnamese businesses. Released under MIT. PRs welcome.