Flowise review

Flowise is an open-source visual platform for assembling AI applications, not just a hosted API or a model. It sits between your application and the models, embeddings, vector databases and tools that power chat assistants and agent workflows. Its canvas lets developers compose single-agent chatbots, retrieval-augmented generation (RAG) flows and coordinated multi-agent systems.

You can run it as a self-hosted server: the homepage gives npm install -g flowise and npx flowise start as the installation and startup commands. Flowise also offers cloud deployment, while its homepage describes on-premises deployment and horizontal scaling with message queues and workers. Applications can call its prediction API or use embedded chat. The architecture can therefore fit as an AI workflow service behind an application or internal tool. The key caveat is the homepage notice: “We're sunsetting Flowise.” Teams evaluating it should establish what that means for their deployment and support needs before adopting it.

The homepage shows an HTTP prediction endpoint at /api/v1/prediction/:id and a Python example that sends a POST request with a question payload. It also lists APIs, an embedded chat widget, and TypeScript and Python SDKs. That gives application developers more than a canvas-only integration: a flow can be called from a backend, or exposed through embedded chat.

For local installation, the homepage documents npm install -g flowise followed by npx flowise start. It also lists support for 100+ LLMs, embeddings and vector databases, though teams should verify compatibility for each component they plan to use. Execution traces can connect to Prometheus, OpenTelemetry and other observability tools.

Documentation depth deserves a check before committing. In one Reddit thread, a Flowise user says the docs are “lacking in detail” and recommends LangChain documentation as a supplement. Check the vendor docs for endpoint authentication, SDK versions and deployment-specific guidance; the homepage does not describe those implementation details.

The hosted Free plan costs $0 per month and includes two flows and assistants, 100 predictions per month and 5MB of storage. Starter costs $35 per month, with unlimited flows and assistants, 10,000 predictions and 1GB of storage; the first month is free. Pro costs $65 per month and includes 50,000 predictions, 10GB of storage, unlimited workspaces and five users. Additional Pro users cost $15 per user per month. Enterprise pricing is custom.

The Self-Hosted (Open Source) option is free under the Apache 2.0 license, with unlimited seats and usage subject to your infrastructure. That removes a per-prediction plan cap, but shifts compute, storage and operations costs to your team. The vendor lists monthly prices; annual billing prices are not published. The vendor does not publish overage rates or API rate limits, so ask how a hosted project behaves after its prediction allowance is reached. The homepage also says Flowise is being sunset, so confirm that hosted plans remain available and supported before budgeting for them.

Flowise describes itself as open source, and its Self-Hosted (Open Source) plan is listed as free under the Apache 2.0 license. The homepage gives npm installation commands and says deployments can run on-premises or in the cloud. It also describes horizontal scaling with message queues and workers, which puts infrastructure planning and operations in your hands.

Self-hosting reduces dependence on Flowise’s hosted service, but it does not by itself settle migration risk. Your application can call a documented HTTP prediction endpoint, while the flows themselves are built inside Flowise. The vendor does not publish a data export policy; confirm how to export flow definitions, credentials and other project data before relying on them. The sunset notice makes that question especially important. Keep model calls, prompts and application logic documented outside the visual canvas where practical, and test a migration path before making Flowise a critical dependency.

Flowise lists execution traces, Prometheus and OpenTelemetry support, on-premises deployment, and horizontal scaling with message queues and workers. Those are useful pieces for an operational deployment, but they do not replace checks for authentication, access controls, data handling, backups and recovery. Confirm the controls and compliance terms that your application requires with the vendor before sending production data.

The hosted plans have prediction and storage allowances, but the vendor does not publish rate limits or overage behavior. For self-hosting, estimate capacity and monitor the infrastructure you operate. The sunset notice is the largest reliability concern: teams should ask about maintenance, security fixes and service continuity, then document an exit route.

Developer feedback is mixed. One Reddit commenter says Flowise is “more conceptually digestible” but “an underpowered system.” Another reports that Flowise documentation lacks detail. These are individual views, not service guarantees, but they point to two practical checks: prototype the full workflow and validate documentation against your production requirements.