Conversational AI Builder

Part of the worlds-biggest-software-project initiative.

An open, LLM-native, no-code chatbot builder that gives product teams the configurability of a developer framework without locking them into a single LLM provider, channel, or pricing model.

Conversational AI Builder is a visual platform for designing, deploying, and analysing LLM-powered chatbots and virtual assistants. It targets product managers, conversation designers, and engineering teams who need to ship intelligent agents faster than raw LLM APIs allow, but with more flexibility than rigid no-code tools provide. The core problem it solves: most existing builders force a trade-off between simplicity (rigid decision trees) and power (developer-only frameworks like Rasa).

Why Conversational AI Builder?

Incumbent pricing escalates fast. Intercom Fin charges roughly $0.99 per resolution on top of $29–$139/seat/month, pushing total enterprise cost into tens of thousands. Landbot WhatsApp plans start at $233/month, Tidio Premium at $2,999/month.
Open-source options are developer-heavy. Rasa requires Python and YAML expertise; Flowise (51,000+ GitHub stars, Apache 2.0) is essentially LangChain with a canvas — primarily developer-oriented despite the visual interface. Botpress is open core, but 68% of its active users are software developers.
Vendor lock-in to specific LLMs. Voiceflow restricts model support to OpenAI and Anthropic on Pro/Business plans, with the free tier locked to ChatGPT. Most SaaS tools cannot switch LLM providers without rebuilding flows.
Workday's 2025 acquisition of Flowise creates uncertainty about long-term open-source commitment in the visual LLM-builder category, leaving room for a community-governed alternative.
Analytics, gap detection, and regression testing are consistently thin across both open and proprietary tools — buyers regularly cite "minimal" or "basic" analytics as a weakness.

Key Features

Visual Flow Building

Drag-and-drop canvas for designing conversation paths, branches, and conditions without code
Blended deterministic and generative nodes — deterministic rules for compliance-critical paths combined with LLM-generated responses for open-ended conversation
Inline LLM testing and simulated conversations within the builder
Template library for common use cases (FAQ bot, lead capture, order status)

LLM and Knowledge Layer

LLM-agnostic integration: OpenAI, Anthropic, Google Gemini, Mistral, and self-hosted open-source models, switchable without rebuilding flows
Retrieval-augmented generation pipeline with document upload, URL scraping, and automatic chunking and indexing
Prompt construction, context window budgeting, multi-turn memory, and tool/function calling managed by the orchestration layer
Model routing for cost control: cheaper models for simple intents, more capable models for complex reasoning

Multi-Channel Deployment

Embeddable web chat widget via a single JavaScript snippet
Publishing to WhatsApp, Slack, Telegram, SMS, and custom REST webhook channels from one agent definition
Webhook support for live data lookups (CRM, order status, booking) mid-conversation
Human handoff to live agents with full conversation context transferred

Analytics and Quality

Dashboards for interaction volume, containment rate, escalation triggers, drop-off points, and CSAT signals
Intent gap detection: automatically surfacing queries the bot fails to answer well and suggesting knowledge base additions
A/B testing across flow variants with statistical comparison
Synthetic conversation generation for stress-testing agent behaviour before deployment
Conversation regression test suites integrated with CI/CD

Governance and Embedding

Role-based access control: builder, reviewer, and admin permissions with audit logging
White-label embedding SDK for ISVs and resellers offering conversational AI inside their own products
PII masking and audit logging for regulated industries
On-premise or private cloud deployment options

AI-Native Advantage

Beyond using LLMs to power conversation, the platform applies AI to the build and operate cycle: scanning conversation logs to identify unanswered queries and draft new KB articles, proposing additional flow branches based on observed patterns, summarising A/B test results in plain language, generating synthetic users for pre-launch stress tests, and scoring human handoff quality to surface coaching opportunities. These capabilities close gaps that today require custom engineering on top of every incumbent tool.

Tech Stack & Deployment

Self-hosted (Docker) or managed cloud deployment, with private cloud and on-premise options for regulated industries
LLM orchestration layer abstracting prompt construction, memory, and tool calling across providers
Vector-indexed knowledge base for grounded responses
REST API and webhooks for full programmatic control over bots, conversations, and messages
TypeScript/JavaScript SDK for custom channels and code blocks; suitable for white-label embedding

Market Context

The no-code AI platform market was valued at approximately $8.6 billion in 2026 and is projected to reach $75.14 billion by 2034, with multi-agent conversational systems growing 327% in less than four months in some tracked cohorts (see research.md). Primary buyers are customer support leaders, product managers, growth marketers, and ISVs embedding conversational AI into their own products. Incumbent pricing ranges from $29/month entry tiers to $2,999+/month enterprise plans and per-resolution fees of around $0.99.

Project Status

This project is in the research and specification phase.
Contributions, feedback, and domain expertise are welcome.

Contributing

We welcome contributions from developers, domain experts, and potential users. See CONTRIBUTING.md for guidelines.

Important: All contributions must be your own original work or clearly attributed open-source material with a compatible licence. Copyright infringement and licence violations will not be tolerated and will result in immediate removal of the offending contribution. If you are unsure whether a piece of code, text, or other material is safe to contribute, open an issue and ask before submitting.

Licence

Licence to be determined. Research recommends Apache 2.0 or MIT to align with comparable open-source projects in the category (Botpress, Rasa, Flowise). See discussion for context.