Patterns
Added supervisor and sequential patterns to enable coordinated multi-agent workflows.
Evaluator Agent
An oversight layer that continuously monitors agent behaviour, validates outputs against defined objectives and constraints, and ensures the agent performs reliably, accurately, and as intended.
Conditional Branching
Enables users to define dynamic execution paths within workflows, including branching logic and iterative loops, allowing flexible control over how agents and tools are executed based on conditions.
Guardrails
Implemented input/output guardrails to enforce validation and safety across agent workflows.
File Storage
File system within the application, allowing seamless file control when using agentflows.
Long-Term Memory (Chat History)
Store chat history to revisit and continue previous conversations across sessions in both Library Agents and the Agent SDK.
Natural-language Orchestration Builder (Agent5i Chat)
Create orchestrations using natural language, with in-chat tools to edit nodes, delete nodes, use for-loop and filter tools, and switch to chitchat mode.
Webhook Support for Agent SDK and Orchestrator
Invoke the Agent SDK and orchestrations programmatically from external systems and applications using webhooks.
Observability platform integration
Integrated an open-source observability platform to enable deeper observability into agent execution, prompts, and responses to monitor and debug agent behaviour.
Model Access Control
Control LLM access at user and group levels so only models explicitly assigned appear in the interface.
User-specific Folders for MCP Servers
Provide user-specific folders in the file_ops and data_transformation MCP servers for better organization and isolation of files and outputs.
Agent Memory & User Memory
Add memory at the agent level using Agent Memory (only that agent will know it) or at the user level using UserMemory (shared across all agents), enabling context retention across interactions and continuity in workflows.
Agent Observability UI
View detailed agent-level execution logs with Observability platform integration under Profile → Traces, making monitoring and debugging easier.
Agent Skills
Extend agents with Skills while editing an agent under the Tools section, enabling modular and reusable capabilities.
Agent Avatars
Personalize agents by uploading custom avatars or choosing from pre-existing icons.
Revamped Orchestration Canvas
Experience a smoother and more responsive orchestration canvas UI.
