August brings Port Workflows to General Availability (GA) as the governed orchestration layer for agents, automations, and approvals, plus new Cursor Cloud Agent syncing and triggering directly from workflows. A wave of workflow, integration, and governance improvements rounds out the month. Here's everything that's new in Port.
Big
Port Workflows is now generally available
image
Port Workflows, the governed orchestration layer for agents, automations, and approvals, is now generally available.
Chain agents, automations, business rules, and human approvals into a single governed flow on shared data. Describe a use case to Port AI in plan mode or start from the visual, low-code builder—trigger the flow via a self-service form or automatically on catalog changes, and trigger GitHub Actions, GitLab, Jenkins, ArgoCD, or your preferred tooling as workflow steps.
Sync and Trigger Cursor Cloud Agents from Workflows
Port syncs Cursor cloud agents and runs into your Context Lake, and workflows can now create or trigger them directly.
See every managed Cursor cloud agent and run them alongside the rest of your software catalog. Specific integration action nodes were added to spin up a new Cursor agent or send a follow-up prompt to an existing one straight from a Port workflow, with optional completion tracking so the run waits until the agent finishes.
image
Medium
Track Claude and Cursor Skills Adoption and Usage
See which Claude and Cursor Skills are actually being adopted, reused, and worth scaling across your organization.
Track invocations, usage depth, active days, and sharing scope to spot Skills that are consistently reused across teams, then promote the proven ones into the Port Skill Registry.
image
Surface Missing Setup Steps Live in AI Chat
Port AI chat now shows a live checklist of the integrations and secrets you still need to connect to make what you built work.
As you build with Port AI, a checklist computed from the conversation tracks any missing integrations or secrets—click an item to connect it right from the chat.
image
Ingest Agent Skills and Plugins from GitHub and GitLab
GitHub Ocean and GitLab v2 integrations now have dedicated skill and plugin resource kinds so you can easily discover and catalog Agent Skills and plugin packages directly from your repos.
Add a
kind: skill
and/or
kind: plugin
resource to your integration mapping and it will discover
SKILL.md
files under common layouts like
.claude/skills/**
and
.cursor/skills/**
, while the plugin detects provider manifests for Claude, Cursor, Codex, and more.
Trigger Azure DevOps Pipelines from Workflows
Trigger and track Azure DevOps pipeline runs directly from a Port workflow, no custom webhooks required.
Add the Azure DevOps action node in Workflow Builder, set the Project and Pipeline ID, and optionally pass a branch, template parameters, or variables—Port queues the run, links straight to it in Azure DevOps, and updates the workflow status automatically as it completes.
Approve and Fill Workflow Inputs from Slack
Responders can now approve or fill out workflow input forms straight from a Slack DM instead of opening Port.
Add a Slack notification target to an Input node and each responder gets a DM that will complement the input node: A simple button-only forms submit on click, Detailed forms opened inside a Slack modal, or Slack link back to the full form in Port.
image
Scope Workflow Permissions to MCP Agent Invocations
Workflow permission policies can now target a dedicated MCP source, letting you control execution by AI agents separately from UI and API callers.
Add a policy rule on the source context using the new MCP value. For example, allow humans via the UI and agents via MCP while blocking direct API calls, or exclude MCP entirely from a sensitive workflow. This lets you restrict which workflows AI agents can trigger, such as blocking MCP execution on workflows with sensitive input nodes or destructive actions, while still allowing human and API access as usual. It gives teams granular control to adopt agentic workflows safely, without an all-or-nothing trust decision about what AI agents can execute.
Manage Workflows as Code with the Port Terraform Provider
Define and manage Port workflows in Terraform so changes go through the same review process as the rest of your config.
Define a
port_workflow
resource with its
node
and
connections
blocks, then run
terraform plan
and
terraform apply
as usual—no more managing workflows outside your review and promotion pipeline.
Browse Blueprints in the New Context Lake Catalog
Context Lake now includes a Catalog view for browsing blueprints, and new blueprints no longer automatically create an Interface catalog page.
Open Catalog in Context Lake to browse a blueprint's entities table and connect data sources, automations, and discovery. Creating a new blueprint no longer adds a catalog table under Interface by default, but you can still opt in via the API or add a page manually.
image
Merge Array Properties from Multiple Sources
Array properties can now merge values from multiple integrations, API callers, or manual edits without overwriting each other.
Enable multi sources ingestion on a string or number array property so each writer owns its own slice—for example, merge vulnerabilities reported by Snyk and Dependabot into one deduplicated list, plus manual edits from the UI.
image
Small
Track Both Identities for Impersonated Workflow Runs
Workflow runs triggered during a "view as" session now record both the admin and the impersonated user.
In the Runs table, the "Who triggered" column now shows "X as Y" for runs started while impersonating another user, and node run details show both "Triggered by" and "Viewed as" for the same attribution at the node level.
image
Skip Notifications for Approvals and Workflow Inputs
Approval steps for self-service actions and workflow Input nodes can now skip sending a notification entirely.
Set
approvalNotification
to type none on a self-service action, or an empty notifications array on a workflow Input node, to keep the approval step while skipping the notification.
Deep-Link Directly to Key Modals via URL
You can now open modals like Credentials or MCP servers directly by adding
?openModal=X
to your organization's URL.
Append
?openModal=X
to your organization home URL—for example,
?openModal=credentials
opens the Credentials modal, or
?openModal=mcpServers
opens the MCP servers modal.
Upload Custom Organization Logos Directly in Port
Organizations can now upload and manage their own logo directly in Port instead of hosting it externally.
Go to Settings → Logo, then upload, drag, or paste a PNG, JPEG, or SVG file up to 5 MB. Delete it anytime to fall back to Port's default logo.
image
Sync Multiple GitHub Organizations from a Single Integration
GitHub Ocean App authentication now auto-discovers and syncs every organization where the App is installed, all from one integration.
Transfer your GitHub App to the enterprise level, install it in each organization you want to sync, and leave
github_organization
unset—Port automatically discovers and syncs every org where the App is installed.
Manage Folders API
Create, update, and delete sidebar folders programmatically now that the Folders API is generally available.
Use the Create, Update, and Delete Folder API routes to manage nested sidebar folders as part of your onboarding scripts or infrastructure-as-code setup.
image
New Filtering and Sorting Options in Mapping
A batch of UX updates adds new filters and sorting to Mapping, plus smaller polish to sync status and the Ask AI button.
Filter Mapping by blueprint or by kind, for a cleaner, more focused view.
image
Navigate Governance with a New Categorized Layout
The Governance section now has a categorized layout with direct links to the Audit Log and Usage Dashboard pages.
Find governance tools faster with a clearer layout, organized into categories with quick links to the pages you use most.
image
Automatically Suspend Integrations After Repeated Sync Failures
Port now automatically pauses scheduled resyncs for integrations that keep failing, alerting admins before resources are wasted.
When an integration fails 3 or more times within a 7-day window, Port suspends its scheduled resyncs and notifies admins to investigate. Admins can trigger a manual resync anytime, and a successful sync automatically reactivates the schedule.
Protect Live Events from Data Loss During Crashes
Live Events now durably queue every webhook so a pod crash or restart can no longer drop events.
Every webhook is now durably written to Redis Streams through a dedicated ingestion service before processing, and any stuck or failed messages are automatically retried. Hosted integrations get this automatically; on-prem integrations can enable it by deploying the ocean-gateway service with a durable Redis instance.
Diagnose and Recover from Mass Entity Deletions
Get an actionable recovery flow when a resync deletes more than half your catalog entities instead of just a warning.
Open the mass-deletion callout on the integration's sync status tab and Port suggests the right recovery action based on the likely cause—run a full resync, roll back to the last mapping snapshot, reapply the current mapping, or open Port AI to diagnose and fix it.
image
Simplify Your UI: Hide Classic Actions and Automations
Ready to go all-in on Workflows? Admins can now give their org a clean, focused experience by hiding legacy actions and automations across the entire UI.
Turn on the toggle in Settings → Actions and Automations to remove classic actions and automations from the self-service hub, entity menus, search, dashboards, the builder, and the audit log. Existing actions and automations keep running via their triggers, the API, and Terraform, and the change is fully reversible.
New Guides
Manage Developer Environments
Governed workflows to create, extend, and tear down developer environments via GitHub Actions, with automatic TTL-based expiry.
Provision Azure Resources with ADO MCP
A Port workflow that reads Terraform modules from Azure DevOps via MCP and has Claude Code open a convention-following PR.
Visualize Claude Skills Adoption
Dashboards showing how Claude Skills are adopted and reused org-wide, and where skill-related spend is concentrated.