Flows

Flows turn a process you have built in Spaces into a single, reusable workflow that anyone can run. You build the logic once, connecting your models, inputs, and outputs, and then publish it as a flow. After that, anyone you share it with just fills in a few fields and gets the output, without ever touching the nodes behind it.

A Flow works with a process of any size, from a three-step image chain to a multi-model production pipeline. If it can be built in Spaces, it can become a flow. Once it is live, you can run it from the Flows gallery, inside Spaces as a node, from a chat assistant like Claude or ChatGPT, or directly through the API.

On this page

• How Flows work
• Create a Flow
• Run a Flow
• Share and personalize Flows
• Tips for building good Flows
• Common issues
• Content rights

How Flows work

Every flow has two sides: the builder who creates it, and the runner who uses it.

The builder designs and maintains the workflow in Spaces. You connect the models, decide what inputs runners will see, set what the output looks like, and choose who can access it. You can share a flow with your team, with a single collaborator, or with the whole world.

The runner does not need to know how the flow was built. You open it, fill in the inputs, and get the output. No nodes, no learning curve. If you want to take it further, you can personalize any flow, which copies the original process into your own space so you can adapt it.

The point of a flow is to remove friction. The complexity stays behind the scenes, and the people running it get the result instead of a tour of how it was made.

Create a flow

Flows are created in Spaces. That is the only place you build them.

Once it is live, you can edit a flow at any time. You can adjust the interface, or update the internal process in Spaces and republish it.

Run a Flow

You do not need to be in Spaces, or even in Magnific, to run a flow. There are several ways to run one, depending on how you work.

Standalone is the fastest way to run. Pick the flow from the gallery, fill in the inputs, and run. This is the mode for runners who need the output without the overhead.

Inside Spaces, a flow becomes a node. When you add a flow to a board it stays fully intact, with its inputs, outputs, and settings. This lets you chain Flows together, so the output of one becomes the input of the next. For example, you can chain a visual generation flow into a variations flow, into a translation flow, into a resizing Flow, and turn a complex process into a clean connected chain.

From a chat assistant, Flows run through MCP. Magnific supports the Model Context Protocol, so you can run any saved Flow from any MCP-compatible assistant, such as Claude or ChatGPT, without writing code. Connect Magnific to your assistant and ask it to run a flow with a given input, and the assistant handles the rest.

From the API, Flows run programmatically. Send the inputs, get a task ID, and retrieve the result when it is ready, or set a webhook and let Magnific notify you.

Share and personalize Flows

Once a flow is live, sharing is one click. The Share button is available both inside Spaces and in standalone mode. You can send the flow to your team, share it with a client, or publish it to the public gallery.

Anyone who can run your flow can also personalize it. Personalizing copies the original process into a new space, ready to edit, adapt, and republish as their own. This is how a single flow grows into shared creative infrastructure that a whole team builds on.

Tips for building a good flow

1. Name your inputs clearly. Runners only see what you label, not the logic. Make every input field self-explanatory so no one needs a walkthrough.
2. Test before you share. Run the flow yourself from the gallery first. What feels obvious to you as the builder may not be obvious to someone running it fresh.
3. Start simple. A three-step chain that works reliably is more useful than a twenty-step one that needs constant maintenance.
4. Use defaults strategically. Pre-filling fields that rarely change reduces friction for the people running your flow.
5. Think in sequences. The most powerful flows are designed to connect. Build with chaining in mind from the start.

Common issues

• The flow produced an unexpected result. First, check your inputs against the labels. Most issues come from providing the wrong kind of input, such as a busy photo when the flow expects a clean product shot. If the inputs are correct and the result is still wrong, contact the flow builder, since the process itself may need updating.
• I cannot find a flow someone shared with me. Make sure it was shared with you or your team, and that your role has access. Access to shared Flows depends on your plan and workspace role.
• A run failed partway through. A flow runs several steps in sequence, so if one step fails, for example a generation is blocked or times out, the run stops. Try running it again. If it keeps failing on the same step, the flow may need to be reopened and fixed in Spaces.
• Running a flow used more credits than expected. A flow uses credits for every generation inside it. A flow that creates multiple outputs across variants, languages, or formats costs the sum of all of them. Check the flow description to see how many generations it runs per execution.

Content rights

Content you create by running a flow follows the same licensing as any other generation on Magnific. What you generate belongs to you, subject to the platform usage terms.