Ai

Dify 1.14.0 Adds Real-Time Workflow Collaboration and HITL API Enhancements

Dify 1.14.0 Adds Real-Time Workflow Collaboration and HITL API Enhancements

Dify 1.14.0 delivers a meaningful product update centered on team collaboration, enterprise workflow control, and platform reliability. The release adds real-time collaborative workflow editing, expands human-in-the-loop support with a service API, and ships a range of MCP, plugin, and OAuth-related fixes designed to improve day-to-day AI application management.

What Changed

The headline addition in Dify 1.14.0 is Collaboration, which allows multiple workspace members to edit the same workflow simultaneously. The feature includes synced graph updates, online presence visibility, and shared awareness of who is working in specific parts of the workflow.

For self-hosted users, the feature is disabled by default and requires explicit configuration. Teams must enable collaboration mode, switch the worker class to a WebSocket-capable option, and define a public socket URL for the deployment.

Another notable addition is a Service API for human-in-the-loop (HITL). This extends HITL beyond console-based interaction and gives developers a programmatic way to build approval, review, or intervention steps into AI workflows.

Dify 1.14.0 also improves its MCP and plugin stack. The release refreshes MCP tool metadata after updates so interface state stays aligned, fixes an MCP server URL issue involving a duplicated /v1 path that could break OAuth flows, and hardens MCP OAuth discovery when malformed JSON is returned. It also corrects schema publishing for checkbox and json_object field types.

On the plugin side, the update includes fixes for auto-upgrade strategy persistence, local installer behavior, file-input handling, and tenant scoping for internal API end-user lookup. These are practical changes that reduce friction for teams managing extensions in multi-tenant or self-hosted environments.

Why It Matters

This version matters because it pushes Dify further into collaborative enterprise AI development rather than solo prompt tooling. Real-time workflow collaboration is especially relevant for product, ops, and AI teams that need shared visibility while designing or debugging complex automations.

The new HITL service API is also important for governance-heavy use cases. It makes it easier to integrate human review into production AI systems, which is valuable for regulated workflows, customer-facing agents, and internal decision-support tools where full automation is not always appropriate.

The MCP, plugin, and OAuth fixes may look smaller on paper, but they address reliability issues that can affect integrations, authentication, and schema consistency. For technical teams deploying Dify in production, these changes help reduce operational friction and improve confidence in the platform’s extensibility layer.

Overall, Dify 1.14.0 is a solid release focused on making collaborative AI workflow building more practical while tightening the platform’s integration and plugin infrastructure.

Official Source: https://github.com/langgenius/dify/releases/tag/1.14.0