Flow sharing between users

[Dexter192]

Being able to share flows between users seems like an essential feature to enable collaboration for the use of langflow in organisations.

I started working on this feature on my fork

The approach that I chose was to add a new table "access_mapping" which grants access to an item (e.g. a flow or a folder) to a target (e.g. a user or later maybe groups). This table can then be referenced e.g. when loading flows or folders

The current state of my fork supports api operations for:

• Sharing folders (including all flows)
• Sharing individual flows (only shared flows of the folder are shown)
• Removal of shared items (flows and folders)

Before proceeding to the frontend implementation of a "share" button, I had another look at prior suggestions regarding collaborative working, I saw that there are existing plans to use Casbin for access control which brought up a few questions:

1. What is the current state of the migration of langflows current functionality to Casbin
2. Does it make sense for me to continue following the access_mapping approach or would it make more sense to read into Casbin and try to adapt the flow sharing approach
3. What is the usual workflow for new features like this. Should I first present a concept/draft as an issue/discussion before starting the implementation?

This discussion likely relevant for @ogabrielluiz & @YamonBot

[YamonBot]

First of all, thank you for working on the flow sharing feature.

I'm one of the contributors who previously proposed adding organization management to Langflow.
At that time, I didn’t suggest real-time collaborative editing — my focus was mainly on enabling user and resource management at the organizational level.

However, from today's perspective, I think organization management alone isn't enough.

In the long run,
I believe Langflow should aim for something similar to Figma or n8n,
where multiple users can open and edit a single flow together in real-time.

With that goal in mind,
organization and role-based management is only the starting point,
and we eventually need to design a structure that supports
real-time collaboration features like resource locking, state synchronization, and conflict resolution.

The current work on access_mapping is definitely
a practical short-term solution.

Sharing folders and individual flows Removing shared items
These are solid improvements at the API level.

However, in terms of long-term architecture,
integrating everything under a Casbin-based access control system
is absolutely necessary.

Casbin will allow us to:

Define fine-grained permissions based on organization, role, and resource types Maintain consistent access control across all resource types Scale more securely and systematically as the platform grows

Personally,
I’ve been looking forward to seeing Casbin integration happen for quite some time.

I'm also very curious about
how the migration to Casbin is progressing on the Core Team's side.

If there’s a shared plan or direction for this,
I'd be happy to collaborate and contribute towards Casbin-based development.

Additionally,
I believe that building a custom, non-Casbin access control system
would only create a steeper learning curve for new contributors and users.
Casbin is already a widely adopted and battle-tested framework,
and relying on it would make it much easier for newcomers to understand and contribute quickly.

That's why I strongly believe
adopting Casbin isn’t just a good option — it’s essential.

[Dexter192]

Thank you for the detailed response and insights!

I agree that Langflow could really benefit from real-time collaboration but this should be a long term goal as you said as well. Being able to share flows/folders seems like a feasible (and neccessary) first step into that direction.

Your comments regarding how essential Casbin is make sense to me and it does seem like a better approach than a custom built solution. I'll happily read into Casbin and collaborate with you for the integration of Casbin into Langflow.

Before we start though, I think we would require some information about the current state from the core team. For instance, what the implications the integration can/may have on the current database/endpoint structure of Langflow as well as to hear their opinion regarding which Casbin model would be the most suitable in the long run.

After reading into it a bit, it seems like RBAC might be the most promising approach. It supports multiple user roles (e.g. viewer, editor, admin). For more fine grained control, we might want to use RBAC with tenants/domains to support roles on a per-project basis.

Edit: One nuisance which highlights the necessety of real time synchronisation is having two tabs with the same flow open. When making changes on tab 1 and switching to tab 2, the changes are simply overwritten with the outdated version in tab 2. This can be rather annoying.

[Dexter192]

@YamonBot @ogabrielluiz
I made a draft PR where I implemented a minimal working example of Casbin into Langflow and validated access: Dexter192#10

I would appreciate some input if the approach that I started with makes sense and whether if I should continue working on it wiht my current approach. I also added one specific question/concern regarding performance in the PR that I would like to discuss