Introduced the Cognigy MCP Server, connecting MCP clients like Claude, Codex, Cursor, and VS Code to the Cognigy.AI REST API. The server enables creation, testing, and iterative improvement of LLM-based AI Agents through a self-improvement loop.
2026.7.0
Cognigy.AI
Released March 31, 2026Features
Generate Scenarios from Conversation Transcripts in the Simulator
Introduced a powerful alternative for generating scenarios based on Flows or AI Agent Nodes in the Simulator. Now, you can create realistic scenarios from your existing conversation data to test your AI Agents with real-world interactions and improve their accuracy. Upload conversation transcripts in CSV format to automatically generate scenarios.New Models by OpenAI and Microsoft Azure OpenAI
Added support for new OpenAI and Azure OpenAI models:gpt-5.4, gpt-5.4-mini, and gpt-5.4-nano, allowing you to choose models based on performance and cost.
NiCE CXone Handover Provider (beta)
Introduced the NiCE CXone handover provider, allowing end users to hand over their conversations from an AI Agent to a human agent in a contact center. This feature is available for Cognigy.AI installations integrated with NiCE CXone.Improvements
- Updated
_.merge()in Code Nodes so the code editor suggestions now match Lodash’s behavior when you combine objects - Added the capability to select a time for Knowledge Connector sync
- Removed the
Not understoodoption from the filter in the Intent Trainer - Moved the Connector Label field to the top of the Add Knowledge window. This improvement makes it easier for users to identify and name knowledge connectors before configuring other settings
- Improved OAuth2 handling for Azure OpenAI LLM connections. Custom headers are now sent with both the LLM request and the OAuth2 authentication request
Bug Fixes
- Fixed the Playbooks schema and description in the Cognigy.AI OpenAPI documentation
- Fixed the issue where the Track Goal Node editor displayed extra headings and corrected the display text
- Fixed the issue with Contact Profile data deletion. Now, selecting Delete Date correctly clears all Contact Profile data and disables the Data Collection toggle
- Fixed the issue where LLMs returned raw JSON strings instead of parsed objects. This issue occurred with some LLM providers when the Response Format was set to JSON object in the Advanced section of the LLM Prompt Node
- Fixed the Overwrite Analytics Node so Input Data and Slots are no longer prefilled with
{}. Leaving these fields empty now keeps existing analytics unchanged - Fixed the issue where field regeneration failed after users added a goal on the New Scenario page in the Simulator
- Fixed the Package import failure caused by Node data containing an invalid
children: nullfield - Fixed the issue where the
<NO_USER_INPUT>placeholder was included in the LLM context even when a real user input existed. Now, user inputs are sent correctly to the LLM
Cognigy Live Agent
Released March 31, 2026Improvements
- Added support for Microsoft 365 OAuth SMTP
Bug Fixes
- Fixed the Conversations Filter API method in the Live Agent OpenAPI documentation:
- Corrected request body structure
- Corrected auth header names
- Fixed
not_equal_tofilter withquery_operatorcausing500 Internal Server Errors
Cognigy Insights
Released March 31, 2026Bug Fixes
- Fixed the
$filterbehavior for the Sessions collection in the Cognigy.AI OData endpoint when values in thegoalsfield were stored astext[]. Now, filters using string functions, such ascontains,startswith, andendswith, work correctly
Infrastructure
Cognigy.AI + Insights + Agent Copilot
Pagination Variable No Longer Required
If you previously added theFEATURE_USE_NEW_PAGINATION_PLUGIN variable in cognigyEnv or extraEnvVars, remove the variable before updating to Cognigy.AI 2026.7.0. This variable is now set by default and no longer needs to be configured manually.