Bot Connect
Bring your own chatbot. You keep the backend; OneMango is the front end — in Ask AI, a pop-up, a full-page tab, or a Home widget.
Overview
Register a customer-controlled chatbot and surface it inside OneMango without rebuilding it. Embed it by URL (iframe) or connect it by authenticated API and let OneMango proxy questions and pass answers back (streamed in Ask AI). Choose where it appears — routed inside the unified Ask AI box, as a pop-up bubble on desktop pages, a full-page tab, or a Home dashboard widget — as a setting, not a separate build. Ask AI routing sends the right questions to the right assistant (yours for its domain, native agents for everything else), picking among several registered bots by their keywords. API endpoint and keys live encrypted in Integrations, never exposed to app code, with optional HMAC request signing.
Register a customer-controlled chatbot and surface it inside OneMango by iframe or authenticated API — routed through Ask AI, a pop-up bubble, a full-page tab, or a Home widget, with keys held encrypted in Integrations.
Highlights
Capabilities
Integration Modes
-
Embed by URL (iframe) — you provide a chat URL
-
Connect by API — we proxy questions to your endpoint
-
Bearer-token authenticated API calls
-
HMAC-SHA256 request signing (verify the call is ours)
-
Multiple external assistants per business
Where It Appears (Placements)
-
Inside the unified Ask AI box (routed by keyword)
-
Floating pop-up chat bubble on desktop pages
-
Full-page tab in the app rail
-
Home dashboard widget
-
Shareable deep link to the full-page assistant
Conversation Experience
-
In-app chat panel for API-backed bots
-
Multi-turn follow-up context (prior turns sent)
-
Markdown-formatted answers (lists, links, code)
-
Streaming answers via Server-Sent Events (in Ask AI)
-
Animated typing indicator while the bot thinks
Ask AI Routing
-
Keyword + topic routing rules per assistant
-
Picks the right one of several assistants by keyword
-
Fall-through to native OneMango agents when off-topic
-
Admin toggle to turn routing on or off
Reliability & Analytics
-
One-click connection test on the Assistants page
-
Automatic health monitoring of API-backed bots
-
Admin alert when an assistant starts failing
-
Usage analytics — questions, share answered, response time
-
Per-assistant and per-placement breakdowns
-
Daily volume chart and recent-activity log
-
CSV export of usage data
Security & Governance
-
Endpoint and keys stored encrypted in Integrations
-
Credentials never exposed to app code
-
SSRF protection (private-network endpoints blocked)
-
Choose which of your own sites may show an assistant's page
-
Group visibility scoping (for the whole app)
-
Admin-only assistant management
Partner AI Agents (A2A)
-
Partner agents read service tickets as an employee you choose
-
Optionally, partners ask Ask AI as that employee and approve the changes it proposes
-
Employees hand a ticket to a partner's read-only skill
-
Verified partner card and per-connection tokens
Limits & Specs
-
Integration modes: iframe or authenticated API
-
Max question length: 4,000 characters
-
Follow-up context sent: Up to 10 prior turns
-
Request timeout: 25 seconds
-
Request signing: HMAC-SHA256
-
Health check cadence: Every 15 minutes
-
Analytics windows: 7 / 30 / 90 days, all time
-
Assistants per business: Unlimited
Use cases
Resources
FAQ
Your backend owns the bot's logic and data. In iframe mode we only render your URL, so nothing passes through OneMango. In API mode we proxy each question and keep a log of the question and the answer, for Analytics and health alerts.
An authenticated HTTPS endpoint that accepts a JSON question and returns a JSON answer (streamed via Server-Sent Events if you want incremental delivery). Endpoint and keys are stored encrypted in Integrations, and private-network addresses are blocked.
Yes. Register as many as you like; Ask AI picks between them by the keywords you assign (topics help only when OneMango's own agents are unsure).
Yes. Run the one-click connection test on the Assistants page when you register the bot to confirm the endpoint answers, then rely on usage analytics (share answered, response time) and automatic health monitoring — admins are alerted when an assistant's questions start failing (checked every 15 minutes).
Anywhere you choose — inside the unified Ask AI box (routed by keyword), as a floating pop-up bubble on desktop pages (one assistant at a time), as a full-page tab in the app rail, as a Home dashboard widget, or via a shareable deep link. Each placement is a setting, not a rebuild.
Assistant management is admin-only, and you can limit Bot Connect — all of its assistants — to specific groups. Endpoints and keys are stored encrypted in Integrations, never exposed to app code, and private-network addresses are blocked.
Yes. The in-app chat panel sends the recent conversation turns with each question so an API-backed bot can answer follow-ups in context, and it renders markdown answers cleanly.
