SaaS Release — 2026-09-14 to 2026-09-27
Window: 2026-09-14 → 2026-09-27
Highlights
Section titled “Highlights”Multi-agent orchestration
Section titled “Multi-agent orchestration”Agents can delegate work to other agents through a first-class orchestration surface. Parent agents can discover eligible agents, start child runs, pass context, inspect child-agent progress and output, and coordinate cancellation so larger tasks can be broken into smaller agent-owned steps without losing visibility.
Agent self-service and platform tools
Section titled “Agent self-service and platform tools”Agents have a broader, governed tool surface for inspecting Kindo state, documentation, conversations, selected tools, agent configuration, runs, and schedules. For organizations enrolled in the Ask before running rollout, tool approvals are also clearer, with approval cards using customer-facing language and summarizing the requested action before a user decides.
Native SAML migration and safer sessions
Section titled “Native SAML migration and safer sessions”Kindo’s SAML sign-in path has moved onto the native authentication stack, with support for IdP metadata import and per-host session handling. The migration removes legacy SSO dependencies, improves session isolation, and ensures API access is revoked when a user or organization is disabled.
Integrations
Section titled “Integrations”Microsoft Outlook and Graph email coverage
Section titled “Microsoft Outlook and Graph email coverage”Microsoft email integrations can inspect more of the mailbox content agents need for investigation and automation. Agents can read and search Outlook mail, inspect message and attachment content, handle forwarded messages and archives, process larger attachments, and answer common mailbox-review questions such as external-recipient activity within a time window.
Expanded security and identity integrations
Section titled “Expanded security and identity integrations”Kindo adds and improves coverage across security, identity, and operations integrations:
- ActivTrak activity, working-hours, directory, and group-management tools.
- Qualys cloud-security posture coverage and migration away from retiring Qualys API paths.
- SailPoint ISC app-deployment and certification-launch flows.
- Abnormal Security message-search reliability and payload alignment.
- Armis connection verification and host handling (early access).
AI control-plane integrations
Section titled “AI control-plane integrations”Kindo adds Hugging Face integration coverage, and organizations with early-access integrations enabled can connect Cursor, Anthropic Claude Enterprise, and Salesforce Agentforce. These integrations expand the data sources agents can use to understand AI usage, administration, and policy controls across connected systems.
Improvements
Section titled “Improvements”Clearer usage and chat-history controls
Section titled “Clearer usage and chat-history controls”Command Center usage charts include an API Usage surface filter so admins can separate public API spend from chat and agent usage. The chat history sidebar keeps single-conversation deletion while removing the bulk-delete mode, reducing the risk of accidental cleanup.
More reliable integration error handling
Section titled “More reliable integration error handling”Integration tools return safer and clearer errors across a broader set of vendor packages. Kindo reduces cases where an integration failure exposes tenant-specific endpoint text, replaces opaque internal errors with actionable validation messages, and refuses invalid vendor responses before they are stored as files.
Reliability fixes for chats, agent runs, and integrations
Section titled “Reliability fixes for chats, agent runs, and integrations”This release includes several customer-observable fixes:
- Chat sessions recover more cleanly when a model produces no first token, when an unsupported model response shape appears, or when a prompt previously caused internal runtime text to leak into the conversation.
- Agent creation, agent-run setup, selected-model changes, and completed step-output viewing are more reliable.
- Persistent workspaces recover better from unusable sandbox states, and file-download tools fail quickly on invalid empty input instead of hanging.
- SSO handoffs, per-host sign-in sessions, and integration connection flows return fewer opaque failures.
- Kimi K2.6 usage has a safer routing path when a provider deployment is unavailable.