May 11

Version: 2.8.0

This release focuses on making agents smarter to build, easier to understand, and more reliable in production.

Knowledge just got stronger

• Improved RAG accuracy and retrieval quality → better grounded responses

• Faster indexing and more stable document handling

• Direct ingestion from:

  • Google Drive

  • SharePoint

  • OneDrive

• Enables smoother workflows like document analysis, intake, and validation

New Launch: Solutions (Beta)

• Ready-to-use Agentic AI solutions to jumpstart use cases

• Designed for faster onboarding and real-world workflows

• Still in Beta as we continue expanding coverage and depth

Agent Builder Enhancements

• Plan before build → define intent and structure upfront for better outcomes

• Flow/plan decomposition → visual breakdown of steps, giving clarity into how the agent thinks and executes

• Checklist (Beta) → preview the exact steps your agent will follow for a given plan, making execution more predictable and transparent

• Faster execution with Plan + Checklist → reduces the time agents spend “figuring things out,” leading to quicker responses, lower costs, and more consistent outputs

• Agent versioning → save iterations, restore anytime, and download previous versions for reuse

• Manage Agent → all configurations (knowledge, integrations, capabilities) in one place for easier control

New Integrations

• Redtail CRM

• Wealthbox CRM

• Alongside existing integrations like Stripe, Google OAuth, Resend, Twilio WhatsApp

• Expands ability to plug agents directly into business workflows

Flow Builder

• Performance and usability improvements shipped

• Still in Preview as we continue evolving orchestration capabilities

Reliability & UX Improvements

• Stronger error handling across builds, agents, and integrations

• Clear, user-friendly error messages instead of raw system failures

• Better handling of edge cases like:

  • API timeouts

  • delayed LLM responses

  • partial failures during execution

• Overall smoother and more predictable platform experience

Agent Runtime 1.2

A sharper, safer, more capable runtime:

• Smarter → powered by newer models like GPT-5.4 and GPT-5.1, with checklist-based planning for known flows

• Safer → prompt sanitization + automatic PII detection & masking

• More capable → generate and execute JavaScript & Python code in sandbox

Also includes:

• Multi-model flexibility → switch between providers seamlessly without rebuilding agents

• Faster execution → skip planning for predefined flows using checklists

• Secure execution layer → run code safely with real outputs

More detailed updates on Agent Runtime 1.2 will be available shortly.

With these updates, the platform moves closer to enabling seamless, production-grade app and agent development at scale.

There’s more coming soon - keep building and sharing your feedback with us.

Happy building!