Skip to content

SaaS Release — 2026-05-11 to 2026-05-24

Window: 2026-05-11 → 2026-05-24

Command Center can now show whether an agent run met its intended objective, separate from whether the run completed technically. Admins can inspect per-step verdicts and reasoning, see objective status in run lists and details, and use audit events for compliance review.

Agent and workflow edits now create version snapshots that can be reviewed later. Builders can inspect historical versions, restore a previous version, and rely on audit logging around restore actions.

Pinned integration connections for shared workflows

Section titled “Pinned integration connections for shared workflows”

Workflow builders can pin specific integration connections to a workflow step, allowing a shared workflow to run with the intended connection rather than silently falling back to the runner’s default connection. Kindo now hard-fails unavailable pinned connections with a clear error and records which connection was used in audit logs.

Kindo added or expanded integration coverage across security, ITSM, DLP, monitoring, and developer operations surfaces, including CrowdStrike Falcon Sandbox, Trellix Endpoint Security, Rapid7 InsightCloudSec, Rapid7 InsightVM, Zabbix, Wazuh, Cyble Vision, Burp Suite, MITRE ATT&CK, Cyberhaven, Armis, Zendesk Support Suite, Freshservice, and ServiceNow trigger support.

Microsoft integrations: Fabric and Entra reliability

Section titled “Microsoft integrations: Fabric and Entra reliability”

Microsoft Fabric and Microsoft Entra ID integrations received fixes for authentication and data-access reliability. This improves agent access to Microsoft-backed identity and data workflows that previously could fail before the agent reached the useful work.

Jira transitions now support workflows that require additional fields, and Splunk Enterprise Security updates now better handle status, owner, and disposition changes. These improvements make Kindo agents more useful in ticket and finding workflows where the target system enforces custom process requirements.

Admins now have more granular user-group controls for agent access, file uploads, personal API key creation, free-form chat availability, and integration credential use. Shared agents can be scoped more precisely, and Kindo now warns users when a shared agent depends on tool connections unavailable to them.

Personal API keys with clearer attribution

Section titled “Personal API keys with clearer attribution”

Non-admin users can now generate personal API keys where enabled by policy. Activity from those keys is attributed to the actual user rather than requiring an admin-owned key for every API workflow.

Command Center gained a user filter for recent agent runs and clearer status tooltips, making it easier for admins to understand who triggered runs and what each status column means.

The Tool Policies dialog now aligns connection dropdowns and badges more consistently, improving readability when managing tool access across multiple integrations.

Kindo now documents and surfaces limits around scheduled agents, helping admins understand scale expectations before configuring large scheduled-agent fleets.

Context-window failures now map to a more precise chat error instead of a generic failure. Users get a clearer explanation when a conversation or request has grown beyond the active model’s limits.

Audio transcription in Chat Actions now initializes correctly in the worker path. Audio and video files that depend on transcription no longer fail silently because the transcription provider was unavailable to the task.

Resolved an issue where editing a workflow step after adding a direct trigger could fail when the step referenced user inputs.

Streaming chat sessions now remain healthier during long-running tool execution. Kindo sends keepalive traffic so the browser connection is less likely to drop while an agent is still working.

PDF uploads no longer fail after reaching storage because of response-parsing errors in the upload path.

Additional light-mode polish resolved invisible containers and dark-mode skeleton states that could make loading or stalled content look like redacted content.