Changelog

New updates and improvements.

Custom Views + Platform Improvements

Custom Views (Data Tables)

Create workspace-level data tables that aggregate task outputs into queryable views. Build custom interfaces on top of Beam without external databases.

  • Create views with columns mapped to agent node inputs/outputs
  • Link views together with Has-Many and Many-to-Many relationships
  • Export data to CSV for use in Excel or Google Sheets

Improvements

  • Error Types configuration added to Auto-run settings — define which error codes trigger retries
  • Linked variable dropdown now shows step identifier for clarity
  • Search functionality added to workspace switcher
  • Copy to clipboard icon added on form fields (task data, step I/O, summary)

Memories & Context Engine

Long-term Memory

Agents can now persist and recall information across conversations using a built-in memory store, enabling continuity without re-prompting.

  • Automatic memory extraction from conversations
  • Scoped recall with relevance ranking
  • Manual memory management in workspace settings

Bug Fixes

  • Fixed intermittent timeout on large file uploads
  • Resolved duplicate webhook events on rapid task completion
  • Corrected timezone handling in scheduled triggers

Integrations Hub & Workflow Templates

Integrations Hub

Browse, install, and configure third-party integrations from a centralized hub. Connect Slack, Notion, Google Sheets, and more with one click.

  • OAuth-based auth flow for supported services
  • Per-workspace integration permissions
  • Activity logs for every integration call

Workflow Templates

Kickstart new agents with pre-built workflow templates. Each template includes configured nodes, sample prompts, and recommended integrations.

  • 12 templates across support, sales, and ops categories
  • One-click clone and customize
  • Community-contributed template submissions

Performance & Reliability Update

Faster Execution

Task execution latency reduced by 40% through infrastructure optimizations and smarter batching of LLM calls.

  • Parallel node execution where dependency graph allows
  • Connection pooling for external API calls
  • Streaming output for long-running steps

Improvements

  • Dashboard load time reduced by 60%
  • Workspace search now indexes node configurations
  • Improved error messages with suggested fixes
  • Added bulk task retry from the task history view