New Features
Multi-Team Workspaces
EKB now supports multi-team workspaces with dedicated team sub-domains. Teams can claim their own sub-domain URL, and users can belong to more than one team at the same time. When a user signs up with a sub-domain or is invited to a team, they are added to the correct team workspace without replacing their original/home team. Existing users keep their original team for billing and default ownership, while also gaining access to any additional team sub-domains they belong to. Users who are part of multiple teams can now switch between workspaces — either by accessing a team directly from its dedicated sub-domain or by continuing to use the main application. Team admins can invite existing or new users into their team. Existing users are added as members of the new team without losing their current team access. New users who accept an invite or sign up through an approved team sub-domain are added directly to that team, while users who sign up outside a team sub-domain continue through the original signup flow.Workflow Templates
Start new work faster with reusable templates that pre-fill title, description, labels, and assignees. Teams can share common workflows without copying old issues.Notion and Airtable Skills
Ten pre-built toolkit skills let you query databases, create pages, and update records in Notion and Airtable without writing custom integrations. Skills for Airtable- Airtable-record-lookup — Finds and retrieves records from natural-language criteria, translating intent into Airtable formulas.
- Airtable-data-entry — Creates records from structured or unstructured input with correct field-type coercion.
- Airtable-record-updater — Finds matching records and bulk-updates fields, with confirmation and safety limits.
- Airtable-pipeline-tracker — Manages a deal or project pipeline — advances stages, flags stalled items, and summarizes health.
- Airtable-report-generator — Queries a table, groups and aggregates by a field, and produces a structured breakdown report.
- Notion-standup-digest
- Notion-meeting-notes-creator
- Notion-project-status-report
- Notion-content-tagger
- Notion-knowledge-base-builder
Queue Depth Monitoring for Super-Admins (On-Prem)
Super-admins can now see how many background jobs are queued over time, making it easier to spot processing bottlenecks before they affect users.Model Deprecation Tracking
The system now tracks when AI models are deprecated and surfaces warnings before they’re removed, so you can migrate proactively. Model providers sometimes retire models without notice. You’ll now see a clear indicator when a model you’ve used is no longer available.Improvements
Firebase to Supabase Authentication in Cloud (Internal)
Authentication now runs through the backend. Sign-in and session management moved server-side for better security and reliability. No visible changes to login flows.Sign in Directly to Your Team’s Workspace
Each team now has its own sub-domain. Sign in once and land on the right workspace automatically, or pick from a list if you’re on multiple teams.Control How Many Steps Each Helper Agent Can Take
Set a per-agent limit on tool calls when one agent delegates to another. Prevents runaway loops and lets you tune cost vs. thoroughness.Mark Form Fields as Optional or Required
Forms now support required and optional field validation. Required fields block submission until filled, giving users clearer guidance.API Tool Steps Now Retry on Transient Failures
Workflows automatically retry failed API calls up to a configurable limit, surfacing errors only after exhaustion. Retries appear in the step UI.Copy Assistant Responses in Autopilot
A new copy button lets you grab assistant responses with one click. Rich text and code blocks preserve formatting when pasted.Slack Bot Can Now Use Tools in Conversations
The Slack bot can call tools during chat when enabled, letting it take actions like searching docs or updating records without leaving the thread.Supported File Types Shown During Knowledge Base Upload
The file picker now shows which formats are accepted before you choose a file, saving a round-trip on unsupported types.Faster, More Reliable Test Suite
Backend tests now run more efficiently and catch regressions earlier. You won’t notice a difference, but deploys are safer.Python Sandbox Improvements
- Files with ampersands and other special characters now process correctly in the Python sandbox without unintended normalization.
- Fixed an issue where Python sandboxes weren’t respecting the configured E2B template, causing execution failures in some environments.
Resolved Issues
- Zip archives now sanitize filenames and metadata — fixed a security issue where malicious zip files could write outside intended directories; also resolved crashes when archive metadata was malformed
- Failed API calls surface their error details instead of silently continuing — execution logs now show what went wrong
- Improved background job reliability by preventing duplicate task execution and ensuring safe acknowledgments
- Enhanced worker autoscaling to better handle traffic spikes and reduce queue backlogs
- Improved worker tracking to accurately manage in-flight tasks, even during worker restarts or crashes
- Long-running workflow executions now retain their full data when the UI polls for updates, fixing a bug where context was lost mid-run
- Knowledge base now accepts
.xlsfiles in addition to modern.xlsxformat - Files beginning with certain character sequences (such as
P1) no longer trigger false format detection errors in the knowledge base and chat uploads - Oracle database connection failures when using Oracle service names or the thin driver are resolved — connection strings now validate correctly, and table reflection works consistently across Oracle configurations
- Fixed a pagination bug that caused some records to be missing from large data-type exports
- Confluence sync no longer fails on empty pages — child pages now import correctly when a page has no body content
- Unicode characters in names, descriptions, and other text fields now serialize correctly across all API endpoints
- The public chatbot widget now remembers your conversation when you refresh or return to the page
- Audio and video files uploaded to knowledge base articles now correctly store their duration metadata, fixing playback controls
- LLM-powered compute columns in Smart Tables now properly track and charge credits for each operation
- Files with Korean names in Knowledge Base now download with their original filename instead of a romanized version
- The chatbot disclaimer text now expands correctly when you resize the chat widget taller
- PDF extraction could fail to locate referenced files during processing — file references now resolve correctly
- Decimal values are properly validated for password length — whole numbers like
16.0work, while fractional lengths are rejected with a clear error - Fixed a parsing error that prevented certain AWS LLM responses from being processed — requests that previously failed now complete successfully