What is Langflow?

An open-source, Python-based low-code builder for AI agents, retrieval-augmented generation (RAG) apps and MCP servers, with a visual drag-and-drop editor. It runs as a Windows/macOS desktop app, a Docker container or Python package you host yourself, and the site also lists a hosted cloud account.

Who it's for

Developers and teams who want to prototype visually and then customize with Python, deploying flows as an API or MCP server; the docs describe it as model- and vector-store-agnostic. The security docs say it is a code-execution platform, so it is not aimed at running untrusted code or isolating many tenants without extra infrastructure.

Pros & cons

Strengths

The GitHub page lists an MIT license, so the self-hosted software has no license fee.
The docs list a visual editor with component nodes, a Playground for step-by-step testing, prebuilt templates, and custom Python components.
The docs say flows can be exposed as an MCP server or connected to external MCP servers, and components and flows can serve as agent tools; the site also lists Flow as an API.
The GitHub page lists install options for Langflow Desktop (Windows and macOS), `uv`/pip, Docker (langflowai/langflow image) and source, for Python 3.10 to 3.14; the docs give a minimum of 2 GB RAM.
The GitHub page lists observability integrations such as LangSmith and LangFuse, and the docs list Kubernetes, Docker, Caddy, Nginx, Google Cloud and Hugging Face Spaces deployment guides.
The site says it is the 'same Langflow' on OSS or Cloud, and lists a free cloud account plus Premier Support and Professional Services.

Limitations

The security docs say Langflow does not isolate users within a process or restrict access to the host filesystem and network, flow access controls are not security enforcement, and tenant isolation must come from your own infrastructure.
The install docs say Langflow Desktop does not include the Shareable Playground or Voice Mode, and Windows installs may need Microsoft C++ Build Tools.
The Docker docs say official images require authentication, so you must set a superuser password before first use.
The pages read publish no price for the hosted cloud beyond a free account, and none for Premier Support or Professional Services; the only cloud terms found are in a June 2024 launch post (hosted by DataStax, free to use), which may be out of date.
The pages read do not say whether customer flows, prompts or data are used to train models, and the docs pages read do not describe telemetry, data retention or hosted-cloud usage limits. Model and vector-database providers you connect have their own prices and terms.

Plans and pricing

Open source (self-hosted): $0. MIT-licensed software you run via Desktop, Docker or Python; you pay for your own hosting and any model or database providers.
Cloud (free account): $0. The site lists a free cloud account for the hosted platform; usage limits and any paid cloud tiers were not found on the pages read.
Premier Support / Professional Services: Contact sales (Reported). The site lists these as options but publishes no prices; the IBM DataStax page points to watsonx.data pricing for paid tiers.

Prices come from the Langflow site, GitHub page and docs; no public paid-plan prices were found. The cloud's details (operator, free tier terms) come from a June 2024 launch post that says it is hosted by DataStax, and DataStax now appears on IBM's site, so the current operator and terms are not confirmed. Confirm on the live Langflow and IBM pages.

Details checked October 2026 against the vendor's own pages. Prices and limits change, so confirm before you buy. Official pricing page ↗

Sources

This profile was compiled from these public pages. We have not run hands-on tests.

Summary

Langflow — at a glance

Try Langflow →

We couldn't find that tool

Langflow

At a glance