Skip to main content

MCP: Advanced Filters and Feedback Management

Release Notes | June 18th, 2026

Written by Zvi Cohen Zion

Key Takeaways

  • Filter by any field: Query work items, portfolio items, and feedback by status, sprint, assignee, labels, and every custom field type - so your AI agent retrieves exactly what it needs.

  • Connect and promote feedback: Link feedback to existing work items or create new ones directly from it - closing the triage loop without any manual steps in the UI.

  • Update feedback: Edit status, importance, labels, and company on any feedback item through the MCP, enabling fully automated triage and enrichment pipelines.


Filter by any Field

Until now, Craft.io's MCP could only filter work items by type, parent item and date range. In practice this meant every query started with a full workspace download and the AI agent filtered what it could fit in its context window. For large workspaces this made sprint planning unreliable, backlog queries slow, and cross-portfolio analysis impractical.

Workspace and portfolio work items can now be filtered by most system fields - status, importance, assignee, sprint, quarter, labels, objectives - and by every custom field type. Feedback items can be filtered by both system fields and form fields. Agents now apply filters server-side, retrieving only the items that match rather than pulling everything and processing locally.

The practical difference: a sprint planning query that previously fetched 500+ items now returns only the items in the active sprint, assigned to the right people, with the right status. Faster answers, smaller context windows, and results that hold up at enterprise scale.

Note that our Craft.io MCP is still on closed beta - join the waitlist to get access.


Connect and Promote Feedback

Until now, the Craft.io MCP could read feedback and read which work items feedback was linked to - but it could not write those connections. An agent doing triage could categorize feedback, identify the right feature to connect it to, and then stop. A human had to open the UI and make the final click.

This release adds two complementary actions:

  1. Connect a feedback item to an existing work item - the connection appears in Craft.io exactly as if you had made it from the app, with full visibility in both the portal and the work item panel.

  2. Promote a feedback item to a new work item - title and description are pre-populated from the feedback and the link between them is created in the same action.

For teams running agentic triage - scanning feedback, grouping by theme, matching to roadmap items, and surfacing new features from unlinked signals - this removes the last manual step. An agent can go from raw customer signal to a linked, structured work item without a human in the loop at any stage.


Update Feedback

Until now, creating feedback via MCP was a one-way action. Once submitted, any edit required opening the feedback portal and making the change manually. This made automated triage pipelines incomplete - an agent could ingest and classify feedback but could not write its conclusions back.

From your AI agent, you can now update a feedback item's title, status, internal status, importance, labels, and company. A complete automated triage flow is now possible end to end: ingest new feedback, classify it by theme and priority, update status and labels, connect it to the relevant feature, and promote it to the roadmap if warranted - with no UI steps required at any stage.

Did this answer your question?