Flowise
Visual builder for LLM agents and RAG. It's for people who write code. It's free.
Get patch notes
Same daily newsletter as the rest of the site. One email a day.
No affiliate link on file. Outbound goes to github.com. Identity checked October 21, 2026. Score null is from the run on October 21, 2026.
Grade
We tried it ourselves. Grade D.
- Grade
- D · We tried it
- Score
- 5/10
- Price
- Free
Best for
Jobs people use this for.
Test details
Visual drag-and-drop builder for LLM agents and RAG flows. Apache-2.0 core, with a separately licensed enterprise folder that was not re-verified. The repo is archived. Paid Flowise Cloud exists.
Hands-on review · 2026-10-21
TestedD5/10- CPU-only box: 8 cores, 15 GB RAM (about 4-10 GB free during tests), no GPU.
- API-driven, not the visual editor
- 4 prompts per backend
Flowise ran on our box and its template agent worked well with a free hosted model, using the calculator and remembering a name. Getting there was rough. The first install hung for 20 minutes, we had to create a local admin login, and we had to switch off a security check before it would talk to a local model. Even then, local Gemma 4 got stuck calling the calculator in a loop. The project has also been archived on GitHub.
Works with a hosted model, but heavy setup, a local tool loop, and an archived repo.
- Coverage
- Self-hosted Tool Agent template: 2 maths questions and a 2-turn memory question, local and hosted.
- What we tested
- Install and run Flowise locally, then build its 'Tool Agent' template (calculator tool plus chat memory) and ask 2 maths questions and a 2-turn memory question. Run with local Gemma 4 and a free hosted model.
Method
npm install flowise 3.1.4 (Node 20) into a local folder; flowise start on port 3311. Flows created and run through the REST API from the marketplace template, swapping the model node to ChatOllama (gemma4-e2b) or ChatOpenAI with OpenRouter's base URL (nemotron-3.5-super-vl:free).
Prompts
- What is 1234 * 5678? Use the calculator tool.
- What is (17.5 * 4) + 12.25? Use the calculator tool.
- My name is Dana. Please remember it.
- What is my name?
Measured
- install_attempt1Hung for 20 min in puppeteer's Chrome download (blocked on this box); killed by our timeout
- install_attempt2136 s with PUPPETEER_SKIP_DOWNLOAD=1, 3.2 GB of node_modules
- startup_s18
- first_run_frictionEvery API call returned 401 until we created a local admin user (stored only in the box's SQLite); the default SSRF policy refused the local Ollama ('Access to this host is denied by policy') until we set HTTP_SECURITY_CHECK=false; telemetry is on by default (we turned it off on restart)
- local.multiplyAgent stopped due to max iterations. 15.5 s, failed
- local.addAgent stopped due to max iterations. 19.0 s, failed
- local.rememberHello Dana! I have remembered your name. How can I help you today? 1.0 s, not scored
- local.nameYour name is Dana. 0.4 s, passed
- local.noteThe calculator returned 7006652 correctly each time, but the agent kept calling it 15 times and stopped at max iterations
- openrouter.multiplyThe product of 1234 and 5678 is **7,006,652**. calculator, 1.4 s, passed
- openrouter.addThe result of \((17.5 \times 4) + 12.25\) is **82.25**. calculator, 12.2 s, passed
- openrouter.rememberGot it, Dana! I’ll remember your name. How can I help you today? 0.6 s, not scored
- openrouter.nameYour name is Dana. 0.6 s, passed
Pros
- With a hosted free model the template agent got 3/3 in 0.6-12 s
- Chat memory worked with both models
- Large catalog of ready templates and nodes
Cons
- Heavy install (3.2 GB) that hung on its first try
- Local model setup needed a security override
- With local Gemma 4 the agent looped on its tool and gave no answer, though the same model worked in AutoGen
- The GitHub repo is archived (Aug 2026)
- Status
- Not run
- Date
- October 21, 2026
- Time to first output
- —
- Cost
- —
- Result
- —
- Evidence
- Log linked below
Queued task. Run a fixed multi-step workflow with one tool call and one failure case. Log setup time and cost.
FAQ
What is the agent-tested verdict for Flowise?
Visual builder for LLM agents and RAG. It's for people who write code. It's free. We tried it ourselves. Grade D.
How much does Flowise cost?
Free. Free, open source self-hosted (paid Flowise Cloud exists)
Do sponsors outrank this tool?
No. A sponsored slot is labeled and sold separately from tiers. It cannot move this listing.