One inbox for every channel.
One AI assistant answering first.
Self-hosted omnichannel customer engagement platform with an AI assistant that reads your company knowledge base to reply in seconds, knowing exactly when to hand off to human agents.
- Unified channels
- 12
- AI providers
- 7
- AI models
- 23+
- Seat licenses
- 0đ
- Data sovereignty
- 100%
channels unified into one inbox
switch anytime via dashboard
pre-configured or custom models
zero per-seat licensing fees
all data hosted on your own servers
Four leaks in traditional customer support
Not every problem needs AI. But all four issues below stem from the same root: fragmented conversations and humans repeating tasks machines can solve.
Messages scattered across seven disconnected channels
Facebook page handled by one agent, WhatsApp by another, email in Outlook, website live chat in a third tool. Nobody has a unified customer overview, and dropped messages happen daily.
Every channel routed into a single queue, one unified customer profile, automatic agent assignment.
70–80% of inquiries are repetitive
Shipping costs, inventory checks, warranty policies, store hours. Agents type identical answers hundreds of times a week, while complex high-value inquiries are left waiting.
AI assistant reads your documentation to answer repetitive queries instantly 24/7, no manual shifts needed.
Off-hours create silence, and customers abandon
A customer messages at 10 PM and gets a response the next morning. For purchasing decisions, ten hours of silence usually means a lost sale.
Flexible scheduling: AI active 24/7 or automatically triggers during off-hours to cover staff gaps.
Rigid chatbots give wrong answers and anger customers
Rule-based bots only match exact keywords. Any slight deviation results in off-topic answers or hallucinations, with no smooth path to a human agent.
AI strictly answers from indexed enterprise documents, admits when uncertain, and smoothly transfers to human agents.
Two layers: workspace for humans, AI handling frontline
The foundation is a full-featured omnichannel support desk: unified inbox, customer profiles, routing, labels, and reports. The top layer is an AI assistant operating frontline, escalating only when necessary.
Full-featured omnichannel desk
Unified inbox, customer and company profiles, auto-assignment, labels, internal notes, canned responses, automation rules, proactive campaigns, CSAT surveys, and detailed reporting by agent and channel.
Assistant trained on your docs
Ingest website, pricing pages, warranty policies, PDF files. The assistant retrieves relevant passages before answering. Each inbox can bind an independent assistant with tailored tone and scope.
Public Help Center
Multilingual, SEO-ready knowledge base enabling self-service before reaching chat. The same source serves both self-service visitors and the AI assistant knowledge base.
Four steps from customer message to resolution
Transparent, fully auditable, and human agents can intervene at any second.
Feed your exact policies, products, and workflows
No hallucinations. The AI reads from verified documents you upload, extracts precise answers, and cites exact references.
This impacts response quality more than model choice
Conflicting or outdated docs produce incorrect replies even with premium models. Our built-in feedback loop lets human agents flag incorrect replies along with the exact source chunks used, making knowledge gaps easy to pinpoint.
Beyond answers — An AI assistant that executes actions
Look up orders, create tickets, check inventory, and send webhooks to your internal CRM.
| Tool | Trigger condition | Observed outcome |
|---|---|---|
| Knowledge Base Search | Questions regarding products, pricing, and policies. | Document-grounded answer with source citations. |
| Human Escalation | Customer requests human agent, or assistant lacks sufficient data. | Conversation routed to human queue, AI steps back. |
| Tag Assignment | Detects intent: complaints, pricing inquiry, warranty, return. | Clean reporting without manual tagging overhead. |
| Priority Elevation | Frustrated sentiment or critical order blocker detected. | Escalated tickets prioritized at the top of the queue. |
| Conversation Resolution | Customer satisfied, or prolonged inactivity following resolution. | Queue stays tidy without unresolved stale conversations. |
| Custom System Webhooks | Live data lookup: order tracking, inventory, booking slots, debt. | AI answers with live data: "Your order is dispatched from Warehouse #3" instead of generic delays. |
Operating schedules
Operate 24/7, during business hours, or strictly off-hours to cover night shifts. Many companies begin by delegating unmanned off-hours to AI.
Auto-close inactive chats
Conversations with no response after a set duration are automatically resolved with a courteous message, protected by safeguards to never close active human threads.
Automated FAQ clustering
Analyzes real conversational logs to extract recurring questions and cluster intents. Once human-approved, they enter the knowledge base so the assistant gets smarter every week.
- Load-balanced auto assignment
- Event-driven automation rules
- Bulk operation macros
- Canned responses
- CSAT satisfaction surveys
- Proactive broadcast campaigns
- Custom customer attributes
- Reports by agent / channel / tag
- Bilingual UI (Vietnamese & English)
The entire stack runs on your private infrastructure
No proprietary SaaS lock-in. You own the keys, the database, and the entire conversation history.
Lighter than you think
A single-core VPS with 3.8 GB RAM runs the entire stack using roughly 1.35 GB. Base memory footprint stays lean; serving more traffic scales incrementally during document embedding.
Baseline specs
Mid-load single tenant: 2 vCPU / 4 GB RAM is comfortable. Multi-tenant setup for 10 light brands: 4 vCPU / 8 GB RAM. Upgrade based on verified traffic telemetry.
Pre-packaged, live in a morning
Docker container deployment, automated custom domain & SSL, automated backup routines. The AI assistant is a pluggable add-on that leaves core files untouched for seamless upgrades.
Two operational models based on who manages infrastructure
Both the support desk and AI assistant operate on open-source foundations with zero per-seat licensing fees. The only difference is where the server is hosted and who manages the AI model billing.
| Self-hosted by you | Managed by Dev00 | |
|---|---|---|
| Best suited for | Enterprises with strict data compliance: fintech, healthcare, education, regulated entities. | Retail chains, marketing agencies managing multiple brands, software resellers. |
| Servers & data | On your infrastructure. Dev00 provides setup and technical support. Chat logs never leave your servers. | On Dev00 managed infrastructure, segregated per client. Zero sysadmin overhead for your team. |
| AI model billing | You connect your direct API keys with model providers and set custom spending limits. | Dev00 manages billing and delivers a consolidated invoice matching exact token consumption. |
| Readiness | Production Ready running live in production. | Near Ready tenant segregation ready; tenant portal in final review. |
The economics for your business
Adjust the sliders according to your real business metrics to compare AI operational cost against human staffing.
Your business metrics
Monthly financial impact
100% deduplicated for private server hosting and actual AI token consumption.
* This calculator provides estimates for financial planning, not a binding quotation. Savings represent unlocked bandwidth for higher-value workflows. In practice, most organizations utilize it to expand service capacity and reduce turnaround time rather than trimming headcounts.
What is live today, what comes next
Full engineering transparency. The platform is running live with real production LLM inference, verified end-to-end, but several roadmap modules remain underway.
| Feature module | Status | Details |
|---|---|---|
| Omnichannel Desk | Live | 12 channels unified into a single queue and customer profile. |
| Autonomous AI Assistant | Live | Verified with production model inference: grounded responses, admits unknowns, executes tool chain (search → tag → prioritize → escalate). Ingests websites & PDFs with strict perimeter security. |
| Tools & External APIs | Live | 5 native tools live; custom HTTP webhooks fully executable, configuration dashboard in final polish. |
| Telemetry & Admin Portal | Live | Full telemetry per turn (model, tokens, latency, tool calls). Five dedicated management screens with bilingual Vietnamese & English support. |
| Multi-tenant Portal | In Progress | Platform-level isolation complete. Tenant role permissions underway. Per-tenant AI key lock and quota limits currently in staging. |
| Copilot for Agents | Planned | Decoupled module, planned for future update upon client request. |
-
WEEK 1
Deploy & connect channels
Server setup, domain & SSL provisioning, channel integrations, team onboarding to unified inbox. Immediate value: zero dropped chats, accurate logs.
-
WEEKS 2–3
Ingest docs & dogfooding
Review documents, ingest website & policy PDFs, fine-tune assistant tone. Staff performs internal verification before customer exposure.
-
WEEK 4
Live rollout, scoped
Activate on a single channel and schedule (typically off-hours: lowest risk, highest immediate ROI). Monitor via feedback logs.
-
MONTH 2+
Data-driven expansion
Review resolution rates and flagged answers to expand knowledge base. Scale across all communication channels and peak hours.
Test on your own company documents
Send us sample documentation (PDF, Word, or web link). We will spin up a live demo with your data so you can verify accuracy firsthand.