Analytics cookies only (Google Analytics, PostHog). No ad networks, no cross-site tracking. See our privacy policy.
Self-host Flowise
updated Nov 2026deployed by usprices checked · Jul 2026Install verified on Ubuntu 26.04 · Nov 2026 · how we test
We earn commissions when you shop through the links below. Full disclosure →
A visual builder for LLM applications and agents: drag nodes onto a canvas to build chatflows, multi-agent flows and assistants, connect them to chat models, vector stores, tools and document loaders, store credentials in the app, and call or embed the result through an API or chat widget. A Node.js server with a React UI that keeps its data in SQLite by default, or in PostgreSQL, MySQL or MariaDB.
Pick Flowise when you want to assemble LLM chatflows and agents on a visual canvas instead of writing the orchestration code: you connect chat models, vector stores, document loaders and tools as nodes, save credentials in the app, and call the finished flow through an API or embed it as a chat widget. It is a Node.js server with a React UI that keeps its data in a local SQLite database unless you point `DATABASE_TYPE` at another one. The licence is Apache-2.0 except for `packages/server/src/enterprise` and a few files carrying their own copyright notice (such as `IdentityManager.ts`), which are under a separate commercial licence. Our snippet is based on the project's `docker/` Compose setup, with the Docker Hub image pinned to a release tag, the data directory bind-mounted and PostgreSQL as the database (see the caveats for why not SQLite). We did not find an upstream RAM minimum, so our 1024 MB floor is an estimate. The catalog's Langflow is the Python alternative with a similar canvas, and Dify is the larger platform with its own knowledge base, workflow and app layers.
What you need
Any VPS with at least 1024 MB of RAM
A domain you control — most self-hosted setups need HTTPS in front of them
Flowise's Docker README and repository licence file flag these before you rely on it:
non-negotiableThe container runs as the non-root node user (uid 1000), so a bind-mounted data directory must be writable by it; the README gives chown -R 1000:1000 for Linux hosts, which our snippet does. The README lists DATABASE_PATH, LOG_PATH, SECRETKEY_PATH and BLOB_STORAGE_PATH as the variables to set to persist flows, logs, credentials and uploads. With the default SQLite database, 3.1.4 crashes at startup (its SQLite session store throws this.db.exec is not a function; we hit it on our test VM, and upstream issue #6688 reports it for 3.1.4 and not 3.1.3), so our snippet uses the documented DATABASE_TYPE=postgres instead and points the other three paths into the data directory; back up both that directory and the db_data volume.
non-negotiablePort 3000 serves plain HTTP. The first visit asks you to register an admin account, and the credentials you then store for model providers and databases sit in that data directory; put a TLS reverse proxy in front before exposing it, and set APP_URL if you do.
non-negotiableThe Compose file in the repository uses flowiseai/flowise:latest and exposes dozens of optional variables for databases, storage back ends, queue mode and SSO. We pin 3.1.4 (Docker Hub tag dated 2026-07-29) and set only the database and persistence paths; the enterprise-only features such as SSO and workspaces are in the commercially licensed directory.
non-negotiableFlows can run custom code and tools. The environment-variable reference covers TOOL_FUNCTION_BUILTIN_DEP, TOOL_FUNCTION_EXTERNAL_DEP and DISABLED_NODES for restricting what a flow may import or use; read it before giving other people editor access.
Flowise used 697 MB of RAM at idle on GCP e2-standard-2 · Ubuntu 26.04 LTS · Docker 29.9.3 (first boot, measured Nov 2026). We list a minimum of 1024 MB.
Is the Flowise install snippet verified?
The Flowise install snippet is marked verified on Ubuntu 26.04 (Nov 2026).