Visual flow builder
Conversation steps are nodes on a canvas, connected by drag and drop with a live preview, so a bot can be built and changed without code.
All projects
Ai Ml
Live in productionConferBot — No-Code AI Chatbot Builder SaaS
ConferBot is a no-code AI chatbot platform we built. Businesses design conversation flows on a visual canvas, add an AI agent that answers from their own knowledge, and publish one bot to their website, WhatsApp, Messenger, Instagram and other channels.
Updated
For ConferBot
- Web app (bot builder and inbox)
- Website chat widget
- Messaging channel integrations
- Marketing site
Overview
Overview
ConferBot lets a business build a chatbot without writing code. Flows are assembled from nodes on a visual canvas or started from a template library, and the finished bot runs as a website widget and on messaging channels including WhatsApp, Messenger, Instagram, Telegram, Slack and Discord. Conversations from every channel arrive in one inbox, where a person can take over.
We built the platform on React, Node.js, Express and MongoDB, with WebSockets for live conversations. It has grown well past a basic bot builder: the public site now describes an AI agent with human handover, a choice of 21 AI model providers on the customer's own API key, a knowledge base, ticketing and analytics, sold as a free plan and three flat-priced paid tiers.
The story in three steps
From the problem to what runs today
01 · Where it started
The challenge
Small and mid-sized businesses get questions on more channels than they can staff: the website, WhatsApp, Instagram, Messenger. Most chatbot tools either need a developer for anything beyond a FAQ, or pick the AI vendor for the customer and fold its cost into per-message pricing.
Read the full challengeShow less
ConferBot needed a builder a non-technical person could finish without help, one place to answer every channel, and a clean way to pass a conversation to a person when the bot should stop.
02 · What we built
The solution
We built the builder in React as a canvas: each step of a conversation is a node, steps are connected by drag and drop, and a preview shows the bot as it is edited. The same flow is published to the website widget and to each messaging channel. Conversations are stored in MongoDB and pushed over WebSockets into a shared inbox, so an agent can pick up any chat with its full history.
Read the full solutionShow less
On top of the flows sits an AI agent. The business gives it a persona, its own knowledge and guardrails (topics it may and may not discuss), and sets when it escalates to a person: on negative sentiment, low confidence, a turn limit or a time limit. The model provider is a setting on each bot, so switching models does not mean rebuilding the flow.
03 · Where it stands
The impact
ConferBot is live at conferbot.com, with a free plan of 600 chats a month and paid plans from $19 a month, priced per workspace rather than per seat or channel.
Read the full impactShow less
We have no support-cost or response-time data of our own for ConferBot's customers, so we quote none.
Chapter 1
How does one ConferBot flow run on every channel?
The idea behind ConferBot is that a business builds a conversation once. A flow is a set of nodes on a canvas: messages, questions, buttons, conditions and integration steps. It is stored once, and the same flow is what runs on every surface the business connects.
Those surfaces fall into four groups on the product's own channel page:
- The website widget, the chat bubble most visitors see.
- Messaging channels: WhatsApp, Messenger, Instagram, Telegram, Slack, Microsoft Teams, Discord and LINE.
- Website builders and store platforms, from WordPress and Shopify to Wix, Webflow and Magento, each with its own install path.
- Mobile apps, through SDKs for React Native, Flutter, Android and iOS.
Because there is one flow, a change to a greeting or a qualifying question is made once and shows up everywhere. And because every channel writes into the same conversation store, every conversation lands in one shared inbox, whichever channel it started on. That inbox is where the bot and the team meet: the same record a person opens is the record the bot has been writing to, so there is no copy to keep in sync.
Chapter 2
How does the bot know when to hand a chat to a person?
Most chatbots fail at the edge of what they were scripted for. ConferBot's answer is three layers, each taking over where the last one stops.
1. The scripted flow
The flow answers common questions and collects context: a name, an email, the reason for the chat.
2. The AI agent
When a question goes beyond the flow, an AI agent answers from the business's own knowledge: its website, documents, Q&A pairs and help articles. The agent works inside guardrails the business sets: topics it may discuss, topics it must avoid, actions it must never take and how long its replies can be. Where the agent can call the business's own systems, any action that changes data can be set to need a person's approval.
3. A person
Escalation rules decide when a person steps in: negative sentiment, low confidence, a maximum number of turns or minutes, or the visitor simply asking. Each escalation can carry a priority from low to critical. The person sees the full transcript, the details the bot collected and any uploaded files, so the customer never repeats themselves.
A person can also step into a chat the bot is still handling. The bot pauses while they are there, and if they hand the chat back, it carries on from the same step. Handed-over chats become tickets that can be assigned, prioritised, labelled, snoozed and closed.
Chapter 3
Why is the AI model a setting rather than a rebuild?
Many chatbot products choose the AI vendor for their customers and fold its cost into per-message pricing. ConferBot does the opposite: the AI model provider is a setting on each bot, the business brings its own API key, and changing provider does not mean rebuilding the flow.
The product's model page groups the 21 supported providers by where they develop and host their models, across the United States, China, France, Canada and Singapore, and also supports running a self-hosted model on hardware the business controls. That grouping reflects four reasons a business actually picks one:
- Cost. On a support bot answering thousands of questions a month, the price gap between providers is real, and with the business's own key the saving is theirs.
- Where data goes. Some teams cannot send customer text to certain countries. Choosing a European or a self-hosted model is a setting, not a migration.
- Outages. Every provider has them. A second provider configured as a fallback means an incident slows the bot down rather than ending the conversation.
- Tone. Models differ more in how they follow instructions than in raw ability, and a business can test its own questions on two providers before choosing.
Keeping the model behind a setting is also what lets the product keep up as models change, without asking customers to rebuild anything.
By the numbers
- 21AI model providers a ConferBot bot can run on
- 13Deploy targets: 8 messaging channels, web widget, 4 mobile SDKs
- 19Native integrations (Zapier counted once)
- 49MCP server tools for AI assistants, across 10 modules
Features
What we built into ConferBot
Template library
Ready-made flows grouped by industry, from booking and scheduling to e-commerce and education, each with a greeting, qualifying questions, a capture step and a handoff that can be reworded and published.
One bot on every channel
The same flow runs as a website widget and on WhatsApp Business, Facebook Messenger, Instagram DMs, Telegram, Discord, Slack, LINE and Microsoft Teams.
Shared inbox with live handover
Chats from all channels land in one inbox; a person can take over from the bot, and chats can be assigned, prioritised and labelled.
AI agent with escalation rules
An AI agent answers from the business's own knowledge inside set guardrails and hands over on sentiment, confidence, turn count or time, passing the full transcript and any lead details.
Choice of AI model per bot
Each bot can run on a different AI provider, from the major model makers, using the customer's own API key.
Knowledge base answers
The bot answers from the business's own help articles and documents rather than only from scripted replies.
Flat-priced plans
A free plan, then Starter, Pro and Business tiers priced per workspace, adding more chats, bots and seats, webhooks and API access, white-labelling and SSO.
Website plugins and mobile SDKs
Beyond the messaging channels, the same bot installs on common website builders and store platforms, and inside React Native, Flutter, Android and iOS apps through SDKs.
Analytics and exports
Conversations, completion and drop-off points, conversion goals and per-channel comparison in one dashboard, with CSV export and integrations to send conversation data on to a CRM.
Real screens · 15
What ConferBot looks like
Mobile app 2
Web 12
Project details
Tech stack
ReactNode.jsMongoDBNLPWebSocketExpress
Links
FAQ
Questions about this project
Can a person step into a ConferBot chat while the bot is still answering?
Yes. An agent can take over any running conversation from the shared inbox. The bot pauses while the person is in the chat, and if the agent hands it back, the bot carries on from the same step of the flow.
Can a ConferBot chatbot hand a conversation to a person?
Yes. The AI agent escalates on negative sentiment, low confidence, a set number of turns or minutes, or a direct request, and the person who picks it up sees the whole transcript and any lead details the bot collected.
Which AI models can a ConferBot bot run on?
Each bot chooses its own provider. The AI models page lists 21 models from the major providers, and teams bring their own API key, so model usage is billed by the provider rather than bundled into per-message pricing.
Which channels can one ConferBot flow be published to?
The website widget and messaging channels including WhatsApp Business, Facebook Messenger, Instagram DMs, Telegram, Discord, Slack, LINE and Microsoft Teams. The same flow runs on each, and replies come back to one inbox.
Related
This project is part of
More work
Other things we shipped
All projects
FeaturedSaasfor Storafleet
Storafleet — Multi-Cloud Storage Control Plane
Storafleet is CountryDevs' own SaaS control plane for object storage. Teams connect buckets they already own on 50+ providers, including AWS S3, Cloudflare R2, Backblaze B2, Wasabi and MinIO, then browse, migrate, search, secure and cost-check them from one console.
Next.js 15ReactRedux ToolkitNode.js+11
FeaturedMobile Appfor Pakka Logistics
Pakka Logistics — On-Demand Goods Delivery Platform
Pakka Logistics is an on-demand goods delivery platform for India. We built all of it: Android apps for customers and drivers, a business portal, an operations console and the website. The web platform is live and both Android apps are on Google Play; the two iPhone apps are in App Store review.
React NativeTypeScriptRedux ToolkitNode.js+12
FeaturedMobile Appfor Slic Solutions Private Limited
LetSlic — Social Media App with Group Chat, Stories and Premium
LetSlic, listed on Google Play as Slic, is a social app for photo posts, 24-hour stories and real-time chat with group moderation. We built it for Slic Solutions Private Limited: the React Native app, two Node.js APIs, the superadmin console and the website, plus Premium and ads.
React NativeTypeScriptReduxNode.js+13
Got a project?
Let's build the
next one together
Tell us what you're trying to ship and which project here is closest to it. We'll reply with the questions that shape the plan, then start with a two-week Discovery Sprint, so you see the screens and the plan before any code is written.
Not ready for a full build? Start with a Discovery Sprint, or get a code audit of an app someone else built. See how a project runs with us.
We work from Bengaluru with clients in India and abroad: app development in Bangalore, for businesses in the UAE and for businesses in the UK.