Kore AI for Process Release Notes

This release introduces enhancements to AI-powered browser automation, workflow features, and general platform improvements. Key updates include the new Delay node, workflow caching, and browser automation enhancements such as configurable auto-healing, model connection updates, expanded model support, hardware upgrades, and improved audit logging.

Workflows

Delay Node for Workflows

A new Delay Node allows users to pause workflow execution at specific steps. Users can configure the node name and define a timeout value, either static or variable-based, between 30 and 86,400 seconds. Workflow changes will be visible in debugging and monitoring tools.

Workflow Cache Management

Workflows now support a built-in cache for storing and retrieving values across steps or runs. Users can read and write cache data in Script nodes, use values in template-supported fields, choose account-level or workflow-level scope, and set custom TTLs. Draft and published runs maintain separate cache data.

Browser Automation Enhancements:

Configurable Auto-Healing

Users can now control LLM-based auto-healing through a dedicated toggle in the Browser node configuration. Auto-healing is no longer enabled by default for new workflows, while existing workflows keep it enabled for backward compatibility.

Model Connection Configuration

Users can now update the connection used by a model within a browser node after the automation is published. While the model itself cannot be changed, the associated connection can be switched to address issues such as expired API keys or rate limits.

Import and Export Enhancements

Browser automation nodes are now included in workflow exports. Imports do not apply auto-mapping, so components must match between environments. Before exporting, users should verify that all required components — such as models, linked workflows, custom scripts and functions, integrations, triggers, agentic apps, OCR models, and automations — exist in the target environment.

Models

Expanded Model Support for Browser Automation

Browser automation now supports a wider set of LLMs across OpenAI, Anthropic, and Google. These additions give users more flexibility and access to the latest model capabilities. Supported models include:

OpenAI:
gpt-4o
gpt-4o-mini
gpt-4.1
gpt-5
Anthropic:
claude-sonnet-4-20250514
claude-sonnet-4-5-20250929
claude-haiku-4-5-20251001
Google (Gemini):
gemini-2.0-flash
gemini-2.5-pro
gemini-2.5-flash
gemini-2.5-flash-lite

Settings

Browser Automation Enhancements:

Browser Automation Playground Locking

The Browser Automation playground now prevents conflicting edits by allowing only one active editor at a time. When a user is editing an automation, all other users enter view-only mode, where editing, running instructions, updating models, saving drafts, and publishing are disabled. The playground unlocks when the active user leaves or after one hour of inactivity.

Enhanced Hardware Options

Users can now select larger hardware configurations to improve performance and increase parallel streaming capacity. In addition to the existing 4 vGPU / 16 GB option, an 8 vGPU / 32 GB configuration is now available. The platform also displays updated streaming limits based on the selected hardware, with credit consumption adjusted automatically.

Expanded Overview Details for Published Automations

The Overview page for published automations now displays detailed configuration information. Users can view key automation parameters, including system prompt, max steps, recording status, auto-healing state, Go to URL value, and file-upload settings, for improved transparency.

Audit Log Enhancements

Audit logs now capture additional account-level and workflow-level events related to browser automation, improving visibility into user actions and automation behavior across the platform.

Original source Report a problem

This release includes a new AI-powered browser automation feature and general improvements across the platform.

New AI-Powered Browser Automation

AI for Process now supports human-like interactions with websites, converting plain-language goals into automated tasks such as clicks, typing, navigation, and validations. The platform securely handles authentication, dynamic content, and errors, enabling end-to-end process automation across web-only and legacy systems.

Key Capabilities

Instruction Generation & Simulation: Turn plain-language goals into step-by-step browser actions.

Reusable Automation Nodes: Secure, reusable workflow steps with retries, timeouts, and visual validation.

Automation Deployment: Safely promote workflows with versioning, approvals, validation checks, and rollback.

Secure & Auditable: Detailed logs, replay visibility, and credential management ensure reliability and compliance.

Original source Report a problem