Bring your flows over.

The visual agent builder, ready for production.

Bring your flows over.

Drag-and-drop agents with real governance and lean context. Open source, Apache-2.0.

Import Flowise JSON opens a chatflow or agentflow on the canvas, node for node. You pick your credentials again, and everything that doesn't carry over is listed below.

• Governance. Every edit waits in a draft while Live keeps serving. Before you publish, replay recent real questions on Live and the draft, side by side. Gate Publish on a test set and roll Live back in one step. Roles, an audit log, and API keys that only run the teams you tick.
• Lean context. Each specialist gets a self-contained task, not the whole chat, and the analyst sees only the tables you tick. Fewer tokens in our internal benchmarks, and every run shows what it used.
• Build agent teams. Describe a team and Build with AI puts a first version on the canvas, with every part and connection checked. Call it from your app with a scoped API key, or put it on a schedule.

Import a flow (2-min guide)Star on GitHub

Node 20+ and Docker

$ npx cuttlely start

or

Clone and Compose

$ git clone https://github.com/cuttlely/cuttlely && cd cuttlely && docker compose up -d

Then open http://localhost:43117, go to Chatflows, and select Import → Import Flowise JSON. Pick a credential on each card and select Save.

The opened canvas after Import Flowise JSON. A short walkthrough will play here; the poster is the bakery helper export.

Two ways in

Pick your path

• One flow at a timeOn Chatflows, select Import Flowise JSON. The wizard checks one chatflow or agentflow export and opens it on the canvas that fits it.
• Just the APIA chatflow still answers on POST /api/v1/prediction/:id with the same request body. Point the call at Cuttlely and use the id it assigns.

Same body, new host

curl -s http://localhost:3000/api/v1/prediction/cf_bakery_helper_7a2c -H "Content-Type: application/json" -d '{"question":"How much are two croissants?"}'
curl -s http://localhost:43117/api/v1/prediction/cf_bakery_helper_7a2c -H "Content-Type: application/json" -d '{"question":"How much are two croissants?"}'

cf_bakery_helper_7a2c is a made-up id. After import, use the id Cuttlely shows on the chatflow. The JSON body does not change.

What carries over

• About 300 nodes, with their settings.
• Connections between the nodes that open.
• Chatflows and agentflows, each on their own canvas.
• A chatflow's prediction API body.
• Custom tools, via Tools → Load.
• FLOWISE_* setting names. CUTTLELY_* aliases win when both are set.

On purpose

API keys, passwords and saved credential ids are removed from the file. Pick a credential on each card, then Save.

Dropped, with a note

The file still opens. The wizard drops Arize, LangFuse, LangSmith, LangWatch, Lunary, Opik, Phoenix, and AssemblyAI configuration cards. Set up tracing in Analytics (Chatflow Configuration → Analytics) and transcription in Speech to Text.

Not yet

• Old embed scripts. Copy the new snippet from the chatflow's Embed tab.
• Whole-workspace exports. Import one flow at a time.
• LlamaIndex nodes.
• Credential matching in the wizard. Select credentials on the canvas after it opens.
• Saved canvas zoom.

What carries over Flowise migration wizard

Before you move

1. Back up the data folder and the encryption key. Keep a copy you can restore if the first import is not what you expected.
2. Export the flows you care about as JSON, one chatflow or agentflow per file.
3. List the credentials those flows use, so you can recreate them under Credentials.
4. Note any website embeds, because the old snippet does not carry over.
5. Try one flow on a laptop first, then move the rest.

What you get after

Nothing is saved until you select Save. The first Save publishes the flow; the next canvas save is a draft. After that, your flow gets governance and lean context, and you can build agent teams around it:

• drafts
• Compare
• gates
• rollback
• roles
• audit
• scoped keys
• teams
• scoped analyst
• tokens per run
• Build with AI

Security advisories we checked

We read the code at v1.3.0 against the six advisories on Flowise 3.1.4 that list no patched release:

• Five don't apply. The chat-message (GHSA-ppmg-4cx6-95hh), upsert-history (GHSA-jvx3-mjpw-r4gh) and credential (GHSA-27w2-26m5-x82c) routes check permissions and team. The BullMQ dashboard isn't mounted (GHSA-rpcc-gw54-mfgx). There's no multi-provider SSO (GHSA-cffm-583c-vffr).
• The sixth, invites activated by an email match (GHSA-vf3j-89vf-r697), is fixed in 1.3.0. A pending invite needs its one-time link. One caveat: someone added with Add User before 1.3.0 was stored as active and can still join by email match, so remove and re-invite them to require the link.

We also checked an older file-write advisory, GHSA-8vvx-qvq9-5948, for Flowise 2.2.7 and earlier. The storage helpers it names sanitize the file name in every storage provider.

This is a code read, not a pen test. If you find something, please report it privately through Security.

Send me a flow that breaks.

Open a new issue and attach a made-up export. Remove API keys, passwords and saved credential ids first.

Open an issue2-min guide

Not affiliated with FlowiseAI. Flowise names and trademarks remain with their owners.