AutomationCAN I VIBECODE
FlowHunt?
BUILD FIRST
- A non-sensitive or read-only version of the core experience
- A manual review step wherever correctness or trust matters
- A narrow companion tool that removes a real piece of work
LEAVE FOR LATER
- Regulated decisions, payments, or irreversible actions
- Claims of correctness, compliance, or uptime that the MVP cannot support
- Sensitive integrations until security, support, and auditability are ready
RESEARCH NOTES
What this tool actually does.
FlowHunt (flowhunt.io) is a no-code platform for building AI chatbots, agents, and automation workflows. Its visual builder is a node-based canvas where users connect blocks — LLM calls, agents, RAG retrievers, web scrapers, OCR, conditions, loops, schedulers, webhooks, custom functions — and configure each via side panels. Agents are built from four primitives: Agents (autonomous workers with a goal, model, and tool belt), Tools (800+ integrations including search, CRM, docs, and APIs), Teams of Agents (hierarchies with a Supervisor planning and specialists executing), and Triggers + Memory (scheduled, webhook, or event wake-ups with conversation memory). The platform includes 120+ pre-built agent templates, knowledge sources (websites, PDFs, videos), multimodal attachments, and model flexibility (leading providers or bring-your-own). Finished flows deploy as embeddable chatbots, REST e
RESEARCHED BUILD PROMPT
Give this to your coding agent.
Build a FlowHunt-style no-code automation studio. Core workflow: a user opens a node-based canvas, drags blocks from a palette (LLM call, agent, RAG retriever, web scraper, OCR, condition, loop, scheduler, webhook, custom function), wires them together, configures each node in a side panel, runs the graph in-canvas with verbose per-node logs showing inputs, outputs, and token spend, then deploys the same graph as three surfaces: an embeddable chat widget, a REST endpoint, and a scheduled/cron trigger. Agents are first-class nodes built from four primitives: goal + model + tool belt; team composition (a Supervisor agent that plans and specialist agents that execute); triggers (webhook, schedule, event) plus shared conversation/persistent memory; and a per-agent credit cap. Include a tool registry with ~10 built-in tools (web search, HTTP request, PDF parse, a stub CRM, etc.) and knowledge-source nodes that ingest website, PDF, and video URLs into a vector store. Frontend: React with React Flow (or similar) for the canvas, a palette sidebar, a node config panel, and a live run console streaming logs over WebSocket. Backend: FastAPI orchestrator with a graph executor, Postgres for run state and history, pgvector for RAG retrieval, and a Redis queue for background agent runs. Deployment layer: generate a REST API from any graph, emit a chat-widget embed snippet, and schedule graphs via a cron service. Do not copy FlowHunt's brand, logo, marketing copy, or UI styling — use your own minimal design. Out of scope: SSO/enterprise auth, multi-tenant billing, 800+ third-party integrations, Photomatic image/video generation, real-time multiplayer collaboration, and production audit compliance. Include a README with setup instructions (env vars, Docker Compose), an architecture diagram, and honest limitations (single-worker scheduling, no distributed agent execution, simplified credit accounting, no versioned graph diffs).