langflow Review

Langflow is a low-code builder for agentic and RAG applications. You build a flow by dragging components (models, prompts, vector stores, tools, agents) onto a canvas and wiring them together, then test it in a built-in playground. Each component is Python under the hood, so you can open and change the code of any of them. A finished flow can be called as a REST API, run as an agent with other components exposed to it as tools, or exposed through Langflow's MCP server so MCP clients can use your flows as tools; Langflow can also act as an MCP client to external servers. The homepage lists integrations including OpenAI, Anthropic, Amazon Bedrock, Azure, Groq, Mistral, Ollama, Hugging Face, NVIDIA, Pinecone, Qdrant, Weaviate, Milvus, MongoDB, Redis, Supabase, Notion, Slack, Gmail, Google Drive and Confluence.

The project is on GitHub under the MIT license and is now part of IBM (the site, privacy links and the IBM Langflow product page are IBM's). You can run it three ways: Langflow Desktop for macOS 13 or later and Windows (download is gated behind a short name and email form), the Python package (uv pip install langflow, Python 3.10 to 3.14), or the official Docker image. The current release is Langflow 1.13, which added audit logging, reusable Google Workspace, Microsoft 365 and Slack connections, and concurrent-editing protection; it also turns off auto-login by default, so new installs must set a superuser password first. The DataStax-hosted version of Langflow on Astra was removed on April 9, 2026, and DataStax points users to Langflow OSS instead.

Langflow itself has no paid plan, seat fee or usage cap. Your real costs are the server you host it on and the model and API providers your flows call (OpenAI, Anthropic and so on), billed by those vendors. The langflow.org homepage still says you can "sign up for a free cloud account", but its sign-up buttons lead to the Desktop download, IBM's Langflow page lists no cloud sign-up, and DataStax removed its hosted Langflow on April 9, 2026. Treat self-hosting or Desktop as the way to run it today. Professional services and Premier Support are offered through the contact form.

Prices checked on the vendor's pricing page on November 10, 2026. Langflow has no separate pricing page; see langflow.org, the IBM Langflow page and the MIT license on GitHub.

What Langflow's own site and docs list:

Langflow suits developers and technical teams who want to prototype RAG pipelines and agents visually, keep the option to drop into Python, and own their deployment. It is a strong fit if you want an MCP server built from your own flows, or if your company already buys from IBM and may later run flows inside watsonx Orchestrate. Because there is no managed hosting from the vendor any more, plan for someone to run, secure and upgrade it.

If you want a similar open-source visual builder with a JavaScript stack, compare Flowise. If you would rather write code than wire nodes, LangGraph or CrewAI give you agent frameworks directly. Teams that want a hosted platform with business-app automations should look at n8n or Relevance AI.

MakerStack's review (a directory funded partly by sponsorships and affiliate links) scores it 8.2/10 and calls it "the most complete free tool for building AI agents and RAG applications visually", praising the absence of paywalls, flows-as-APIs and Python customization; it warns that upgrades and complex flows can be buggy, large canvases get hard to manage, and production use takes real engineering now that the DataStax cloud is gone. SelectHub lists fast low-code prototyping and reusable components as strengths, and stability concerns, dependence on external APIs and limited production readiness as drawbacks. The common thread: reviewers like it for building and testing ideas quickly, and say to budget engineering time before relying on it in production.