Robbie answers can now point back to concrete workspace evidence through sanitized provenance anchors, clickable evidence chips, ignored-provider hints, and compact operational timelines, without changing the single AI request flow.
Changes in this release
- The investigation summary now builds a provenance model with evidence anchors for files, selections, terminal output, Source Control, Docker, Database Explorer, Deployments, Metrics, Connections, and WordPress Toolkit
- Robbie replies can now show a compact 'What Robbie Used' section with clickable evidence chips under the answer
- Lower-confidence and investigative answers can expand into ignored-provider hints and a concise chronological provenance timeline
- Provenance targets are permission-aware, secret-like paths are redacted, and unsupported destinations are shown as disabled rather than exposed
- The internal debug panel now previews provenance anchors, navigation targets, disabled-target reasons, redaction notes, and timeline grouping
[Read the full blog post →](/blog/robbie-v2-provenance-ux)
*Released 4 July 2026 · Category: AI*