Langflow
Type: low-code · Vendor: Langflow / DataStax (IBM) · Language: Python (visual) · License: MIT · Status: active · Status in practice: mature
Provide a Python-based, MIT-licensed visual builder for AI agents and workflows where components plug into a flow on a drag-and-drop canvas, with first-class agent-tool wiring, multi-agent orchestration, and one-click MCP-server publication.
Description. Langflow is an open-source (MIT) visual platform for building AI agents and workflows. Its README states: 'Langflow is a powerful platform for building and deploying AI-powered agents and workflows.' Flows are 'functional representations of application workflows' composed of components where 'each component is a single step in the workflow.' The Agent component 'is critical for building agent flows' and 'provides everything you need to create an agent, including multiple Large Language Model (LLM) providers, tool calling, and custom instructions.' Multi-agent flows are realised by setting an Agent component to Tool Mode and attaching it as a tool to another Agent. Langflow flows can be deployed as MCP servers, turning them into tools for MCP clients.
Agent loop shape. A Langflow flow is a directed graph of components on a visual canvas. The Agent component contains an LLM, tools, and custom instructions, and 'uses LLMs as a reasoning engine to process input, determine which actions to take to address the query, and then generate a response.' Tools are attached by connecting any component's Tool output to the Agent's Tools input. Multi-agent shape: any Agent component can be flipped into Tool Mode and used as a tool by another Agent — Langflow's documented mechanism for multi-agent flows. Flows can also be Run-Flow-attached to an agent or published as MCP servers; observability is wired through LangSmith / LangFuse.
Primary use cases
- visual building of agent flows with multiple LLM providers and tool calling
- multi-agent orchestration by attaching agents as tools to other agents
- RAG flows over vector store components
- deploying flows as MCP servers consumed by external MCP clients
- deploying flows as APIs or exporting them to Python apps
Key concepts
- Flow → visual-workflow-graph (docs) — Functional representation of an application workflow on a visual canvas.
- Component (docs) — A single step in the workflow; configurable and connectable to other components.
- Agent component → tool-use (docs) — Reasoning + tool-calling unit with LLM, tools and custom instructions.
- Tool / Tool Mode (docs) — Any component (including another Agent) can be used as a tool by an Agent.
- MCP server publication → mcp (docs) — Deploy a flow as an MCP server so MCP clients can call it as a tool.
Patterns this low-code implements —
- ★★Agentic RAG
Vector-store and embedding components are documented as a 'common' shape in the flow alongside language model components; multi-agent orchestration capability explicitly includes 'conversation manage…
- ★★Tool Use
Agent component integrates 'tool calling' and 'custom instructions' as core capability; any Langflow component can be attached as a tool via the Agent's Tools input.
- ★★Model Context Protocol
Langflow flows can be deployed as MCP servers and consumed as tools by MCP clients (e.g. Claude Desktop, Cursor). MCP Tools component also exists for the reverse direction.
- ★★Visual Workflow Graph
Flows are explicitly 'functional representations of application workflows' composed of connected components on a visual canvas.
- ★★Structured Output
Langflow ships a dedicated Structured Output component that uses an LLM plus an output schema to transform input into JSON or Table data, with field types str/int/float/bool/dict.
- ★★Supervisor
Multi-agent in Langflow is realised by setting another Agent component to Tool Mode and attaching it as a tool. README also references 'Multi-agent orchestration.' Not framed as a named Supervisor ab…
- ★★ReAct
Original stub claimed ReAct as first-class. Langflow Agent component docs describe an LLM-as-reasoning-engine loop but do not name ReAct explicitly. Downgraded.
- ★★Fallback Chain
Original stub claimed fallback-chain as first-class. Langflow language-model docs mention provider/model alternatives via Bundles but do not document a named fallback-chain primitive. Verified 2026-0…
Neighbourhood
Click any neighbour to follow the lineage. Scroll to zoom, drag to pan.
Alternatives & relatives
Listed as alternative by (4)
References
Provenance
- Last analyzed:
- Last updated:
- Verification status: needs-verification