Flowise
Type: low-code · Vendor: FlowiseAI · Language: TypeScript (visual) · License: Apache-2.0 · Status: active · Status in practice: mature
Provide an open-source TypeScript visual builder — 'Build AI Agents, Visually' — that assembles LangChain-JS-style chains, single agents and multi-agent supervisor/worker systems on a drag-and-drop node canvas, with first-class document stores for RAG and a pluggable tool ecosystem.
Description. Flowise is FlowiseAI's open-source (Apache-2.0) visual builder for LLM applications. Its GitHub tagline is 'Build AI Agents, Visually.' The documentation distinguishes three product surfaces: Chatflow ('designed to build single-agent systems, chatbots and simple LLM flows'), Assistant ('the most beginner-friendly way of creating an AI Agent'), and Agentflow, which is 'the superset of Chatflow & Assistant. It can be used to create chat assistant, single-agent system, multi-agent systems, and complex workflow orchestration.' AgentFlow V2 'represents a significant architectural evolution, introducing a new paradigm in Flowise that focuses on explicit workflow orchestration,' adding typed Agent nodes, Document Store retrieval, conditional and LLM-based router nodes, and an explicit Supervisor/Worker multi-agent shape in V1.
Agent loop shape. A Flowise project is one of three flow types. A Chatflow is a single-agent LangChain-style chain. An Assistant follows instructions, uses tools and retrieves from uploaded files. An Agentflow (V2) is an explicit workflow graph of typed nodes — Agent (an LLM with tools / Document Stores), Condition (deterministic branch), LLM Router (semantic branch via 'Scenarios' and natural-language 'Instructions'), Tool, and Retrieval. AgentFlow V1's Multi-Agent shape composes a Supervisor agent that 'analyzes user requests, decomposes them into sub-tasks, and assigns these to specialized worker agents' over connected Workers, sequenced one task at a time. Document Stores centralise ingestion, splitting and vector-store upsert.
Primary use cases
- no-code/low-code building of chatbots and single-agent RAG apps
- multi-agent supervisor/worker systems on a visual canvas
- RAG via centralised Document Stores with multiple vector store backends
- LLM-driven conditional routing inside an AgentFlow
- self-hosted on-prem deployment
Key concepts
- Chatflow (docs) — Single-agent / chatbot / simple LLM flow on a visual canvas.
- Assistant (docs) — Beginner-friendly AI assistant that follows instructions, uses tools and retrieves from uploaded files.
- Agentflow (V2) → visual-workflow-graph (docs) — Superset of Chatflow + Assistant supporting single-agent, multi-agent and explicit workflow orchestration.
- Agent node → tool-use (docs) — Autonomous AI entity that reasons, plans and uses Tools and Document Stores.
- Document Store → agentic-rag (docs) — Centralised data layer for upload, split, prepare and vector-store upsert.
- Multi-Agent (Supervisor / Worker) → supervisor (docs) — AgentFlow V1 multi-agent shape: Supervisor decomposes and delegates tasks; Workers execute specialised functions.
Patterns this low-code implements —
- ★★Agentic RAG
Document Stores are a centralised RAG layer; AgentFlow V2 Agent nodes query Document Stores by semantic similarity, and Document Stores explicitly support upload, split, vector-store upsert and simil…
- ★★Tool Use
AgentFlow V2 Agent nodes are authorised to use a set of Flowise Tools or MCP tools; the LLM decides a sequence of actions and which tools to invoke.
- ★★Supervisor
AgentFlow V1 Multi-Agent ships an explicit Supervisor/Worker abstraction — the Supervisor orchestrates workflow, decomposes requests, and delegates one sub-task at a time to specialised Workers.
- ★★Visual Workflow Graph
AgentFlow V2 explicitly frames itself as a paradigm shift toward 'explicit workflow orchestration' on a visual node canvas; Agentflow is documented as superset of Chatflow + Assistant.
- ★★ReAct
Flowise ships dedicated ReAct Agent LLM and ReAct Agent Chat integrations under LangChain agents, each explicitly using 'Reasoning and Acting' logic to decide what action to take.
- ★★Structured Output
AgentFlow V2's LLM Node provides a JSON Structured Output parameter that instructs the LLM to format output according to a specified JSON schema. Standalone Structured Output Parser node also exists…
- ★★Fallback Chain
Original stub claimed fallback-chain as first-class. Re-verified 2026-05-24: the Flowise docs path for multi-model fallback returns a 404, no AgentFlow V2 page names a fallback-chain or automatic alt…
Neighbourhood
Click any neighbour to follow the lineage. Scroll to zoom, drag to pan.
Alternatives & relatives
- low-code · visual builder
Langflow
Python equivalent — visual LangChain builder.
- low-code · visual builder
n8n
General-purpose workflow automation with embedded AI Agent nodes; broader workflow scope, less LLM-app-focused.
- low-code · visual builder
Dify
Open-source LLM-app platform with deeper RAG product surface.
Listed as alternative by (4)
References
FlowiseAI/Flowise README
“Source code in this repository is made available under the Apache License Version 2.0”
Flowise docs landing
“Agentflow is the superset of Chatflow & Assistant. It can be used to create chat assistant, single-agent system, multi-agent systems, and complex workflow orchestration.”
Provenance
- Last analyzed:
- Last updated:
- Verification status: needs-verification