June 2026
Releases Shipped
Section titled “Releases Shipped”| Version | Released |
|---|---|
| v5.7.1 | 2026-05-28 |
- Multi-agent AI assistant — the assistant now coordinates specialized sub-agents behind a single orchestrator, producing more accurate, better-reasoned answers to your data questions.
- Visual explanations in chat — the assistant can generate diagrams and inline charts to illustrate its answers, so complex results are easier to understand at a glance.
- AI workflow step — call a large language model directly as a step in your workflow, with multiple providers and access tiers supported.
- Chat on mobile — talk to the AI assistant from the mobile app.
- 140+ new data connectors — a major expansion of supported data sources, with guided connection setup and automatic source discovery.
- Incremental, resumable syncs — syncs pick up where they left off and merge new records by key, so you only move what’s changed.
- Faster Snowflake imports — server-side unload speeds up large imports from Snowflake.
- Forecast workflow step — launch forecasting directly as a workflow step.
- Solver workflow step — run optimization logic as part of a workflow.
- Concurrent macro execution — macro steps can run work in parallel for faster processing.
- Compare and merge projects — see the differences between two projects and safely merge changes between them.
- “Where Used” impact analysis — quickly see where agents and external data connections are used before you change them.
- Allocation root-cause tracing — new lineage and tracing tools make it easier to understand and troubleshoot how allocation results were produced.
- Workflow run history and heatmap — a new run-history view with an activity heatmap to spot patterns and problem runs at a glance.
- Collaborative Visual Workflow Designer — a visual view of your workflow with step color-coding, a redesigned canvas, and live collaboration so teammates can work on a workflow together.
Changed
Section titled “Changed”- More relevant AI answers — improved retrieval pulls in the right supporting context when the assistant answers, and expression-authoring help is now validated against the real runtime to catch mistakes earlier.
- Clearer AI activity — a processing indicator shows while the assistant works on longer queries instead of leaving you with silent dead air, and conversations are now organized by project.
- More robust file imports — improved Parquet handling and clearer, more reliable error reporting for problem files, including malformed XML and tricky text encodings.
- Smoother dashboards and user sync — dashboard rendering fixes and tighter synchronization of users and roles with the analytics layer.
- More resilient connections — self-healing for the caching layer and background-processing fixes reduce stalls and improve overall responsiveness.
Security
Section titled “Security”- Passwordless access for the PlaidCloud support team — all support team members now sign in without usernames and passwords, eliminating password-based credentials as an attack vector for support access to the platform.
- Stronger per-tenant isolation — each tenant now uses its own dedicated signing secrets for sessions and workflow steps.
- Cross-site scripting (XSS) hardening — dialog content and on-screen notifications are safely escaped by default, with HTML allowed only where explicitly intended.
- Hardened file imports — malformed uploads are handled safely instead of surfacing confusing errors.
- Continuous vulnerability scanning — application images are automatically scanned for known security vulnerabilities as part of every build.