Skip to main content

Release 0.5.18 — MCP Credentials, Workflow Status Tracking, and Permission UI

Sri Aradhyula
Sri Aradhyula

Released: 2026-06-22 Chart: oci://ghcr.io/cnoe-io/charts/ai-platform-engineering:0.5.18 Previous release: 0.5.17

Highlights​

0.5.18 makes credential management for MCP tools a first-class experience: you can now pin an OAuth provider connection directly to a custom MCP server so your credentials travel with the tool. Workflow runs now return a direct link so you can track status, and a new permissions panel shows who can access each MCP server. Non-admin users can also save and run workflows without needing global agent access grants.

What's New​

MCP Credentials​

  • Pin provider connections to MCP servers — when adding a custom MCP server, you can now attach an OAuth provider connection (GitHub, Jira, Confluence, etc.) directly to it. The platform forwards those credentials automatically every time the tool is called, so you don't need to re-authenticate per session
  • Credential profile and OAuth lifecycle — the secrets dialog shows a richer view of your connected provider accounts, including connection status and a clear flow for authorizing, refreshing, and revoking tokens
  • Fail-closed OAuth enforcement — if a caller-scoped MCP tool requires OAuth credentials and none are present, the request fails with a clear warning in chat rather than silently proceeding without authentication

Workflows​

  • Live status link in responses — when an agent runs a workflow, its response now includes a direct URL to the workflow run so you can click through to track progress, view logs, and see the final result
  • Real-time status polling in chat — the chat interface polls workflow run status automatically and updates the conversation as the run progresses, so you don't need to navigate away to see when it finishes
  • Non-admins can save and run workflows — team members can now create, edit, and run workflows scoped to their team without requiring a platform-wide agent access grant. Ownership and run permissions are checked at the team level

MCP Server Management​

  • Permissions panel — a new panel in the MCP server settings shows which teams and users have access to each server, and lets admins manage grants directly from the server detail view

Access Control​

  • Platform MCP and agent grants reconciled on startup — the platform now reconciles authorization policy entries for built-in MCP servers and agents when it starts, automatically repairing any drift without manual intervention

Bug Fixes​

  • Workflow visibility for view-only members — users with view-only workflow access can now read workflow configurations at the API level without triggering a permission error
  • Team workflow owners can run their workflows — a user who owns a workflow through team ownership but is not a direct team member could previously be blocked from running it; this is now fixed
  • MCP OAuth callback — the OAuth callback handler no longer crashes when user metadata is absent from the session
  • Popovers clipped at viewport edge — dropdowns and popovers that appear near the bottom or right of the screen now flip direction automatically to stay within the visible area

Breaking Changes​

No breaking changes. Drop-in upgrade from 0.5.17.

Upgrade​

helm upgrade ai-platform-engineering \
  oci://ghcr.io/cnoe-io/charts/ai-platform-engineering \
  --version 0.5.18 \
  -f your-values.yaml

No Helm values changes between 0.5.17 and 0.5.18. No data migrations required.