> For the complete documentation index, see [llms.txt](https://docs.aisera.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.aisera.com/product-release-notes/2026/aisera-product-release-notes-september-17-2026.md). # Aisera Product Release Notes - September 17, 2026 ## Conversation ### What's New #### MCP Client The Aisera platform now supports registering and using external MCP (Model Context Protocol) tools directly within AiseraGPT applications. Once registered at the tenant level and assigned to an application, MCP tools are available for AI fulfillment across all supported channels, including Webchat, Copilot, Slack, and WebEx. Tool results are interpreted by a large language model (LLM) to return conversational responses to end users. * Register an MCP server under tenant settings by providing the server URL and configuring authentication. OAuth, Bearer Token, and no-auth connections are supported. * Select which tools from the registered server are available for the tenant, then assign specific tools to individual AiseraGPT applications under **External Tools** on the application details page. * Adjust fulfillment priority for MCP tools relative to other fulfillment types under **AiseraGPT > Configure > Advanced**. [MCP Client](/aisera-platform/what-can-my-bot-do/mcp-client.md) #### Limitations MCP Client does not support tool calling from Flows or Hyperflows in this release. MCP Client does not support mid-process elicitation or long-running processes in this release. MCP Client does not support user-level access control. All registered MCP tools are available to all AiseraGPT channel users, and all tool calls use a shared tenant-level credential. The WebEx input form does not render consistently for MCP tools with more than 6 parameters. ## Flows ### What's New #### Structured Hyperflow prompt editor The new **Hyperflow Prompt Editor** replaces the single freeform prompt field with a section-based authoring experience. Agent behavior is always defined through dedicated sections: Persona, Responsibilities and Scope, Execution Plan, Completion Criteria, and Example Interactions, which Hyperflow compiles into a prompt configuration using standardized prompt engineering practices. Optionally, you can describe the agent's intended behavior in plain language to generate an initial structured configuration. The new editor is controlled by the **Enable Guided Agent Authoring** feature flag, disabled by default, and can be enabled under **Settings > Configuration > Hyperflows**. * The **Generate** action creates or regenerates the full prompt configuration from a plain-language behavior description. Regenerating replaces all existing section content. * The **Optimize** action refines content across all sections while preserving their intent. An undo option is available after optimization. * A read-only **Preview** combines all sections into a single prompt view and can be copied in Markdown format. * Legacy Hyperflows continue to open in the existing editor. The new editor applies only to Hyperflows created after the feature flag is enabled. Hyperflows retain the authoring experience they were originally created with, regardless of subsequent feature flag changes. [Structured Hyperflow Prompt Editor](/aisera-platform/what-can-my-bot-do/ai-automation-optimize-flow/hyperflow-studio/structured-hyperflow-prompt-editor.md) #### Fetch File and Store action The new Fetch File and Store action downloads a file from any accessible HTTP or HTTPS endpoint, validates it, and uploads it to object storage as an attachment. The returned attachment reference uses the same format as a User Input File Attachment, so it works directly with downstream actions like Parse PDF without additional transformation. [Fetch File And Store](/aisera-platform/what-can-my-bot-do/ai-automation-optimize-flow/ai-workflow-studio/building-workflows/using-workflow-actions/frequently-used-actions.md#fetch-file-and-store) #### Parse PDF action The new Parse PDF action extracts text content from up to two PDF attachments using the tenant's configured PDF parser. Returns plain text by default or structured HTML with preserved headings, paragraphs, and tables when enriched output is enabled. If one file fails to parse, the other's result is still returned alongside a per-file error map.\ \ The platform currently uses Microsoft Form Recognizer as its parsing service. You must purchase and deploy your own Microsoft Form Recognizer instance before enabling this feature. Aisera does not provide the service, and any associated costs are billed directly to you. To enable, select [**Microsoft Form Recognizer (Additional charges may occur)**](/aisera-platform/tenant-setup/aisera-platform-configuration/tenant-configuration-settings/parser/microsoft-form-recognizer.md#microsoft-form-recognizer-additional-charges-may-occur) in **Parser Configuration**. Contact Aisera for assistance with setup and configuration of the service. [Parse PDF](/aisera-platform/what-can-my-bot-do/ai-automation-optimize-flow/ai-workflow-studio/building-workflows/using-workflow-actions/frequently-used-actions.md#parse-pdf) ### What's Changed **Hyperflow V2 is now generally available** Hyperflow V2 is now the default experience for all tenants. The feature flag previously used to control the rollout has been removed, and no tenant-level configuration is required. Review the behavior of your existing Hyperflows after this migration. The updated model may produce different outputs, and some Hyperflows may require prompt tuning to maintain expected behavior. ### Fixes Fixed an issue where cloning a tenant-level Workflow to application scope failed with a "bot ID required" error because the clone process incorrectly validated for a bot association on the source Workflow. Tenant-level Workflows can now be cloned to application scope successfully. Fixed an issue where campaign Workflows incorrectly appeared in the Workflow registration list for Aisera 2.0 applications, causing system-generated names and IDs to appear as selectable options during bot configuration. Fixed an issue where importing a Hyperflow into an environment where a Hyperflow with the same name already existed could create a duplicate instead of updating the existing one. Imported Hyperflows are now always created as draft versions, and imports into an environment with a matching name add the content as a new draft version of the existing Hyperflow rather than creating a separate entry. Fixed an issue where restoring an archived Workflow version could result in input and output variables not being fully synchronized with the restored definition, causing outdated or incomplete variable information to persist until a subsequent save. ## Ingestion and Connectors ### What's New #### **Integration Health Indicators on Detail Pages** Every Data Source depends on an underlying Integration for its credentials. If that Integration has a credential issue, is expiring, or has expired, the Data Source will eventually fail to sync — but the Data Source detail page currently gives no indication of this. Users only find out when a sync breaks. Now the **Integration Detail** page displays a badge with four states — Credential Issue, Expiring Soon, Expired, or Healthy. This includes a banner with more details (issue count/dates, or expiration date and days remaining) and an action button to fix the credential. **T**he same badge and detail banner are displayed on the **Data Source Details** page, reflecting the underlying Integration's status. This section also includes an action link that takes you to the credential information for the Integration. **Important:** the health is related to the linked Integration, not the Data Source configuration parameters. [Integration Status](https://docs.aisera.com/aisera-platform/adding-data-to-your-tenant/integrations-and-data-sources/creating-a-connector/adding-and-managing-integrations-in-your-tenant#integration-status) ### Fixes Fixed an issue where customers were receiving the error, ""Cannot have more than 255 characters." if they were integrating with systems that generate long URLs, to integrate with systems that generate long endpoint URLs (such as ServiceNow or Power Automate). The fix increases the limit for the API URL field so that your can integrate APIs that include URLs with more than 255 characters. Fixed an issue where Slack citations were being returned with just a placeholder when the Slack Bolt SDK was enabled and a request was made for a KB query that generated a response with source citations. Now All citation references are linked to valid sources, with corresponding source details displayed in the "Relevant Documents" or references section. ## Knowledge Base Article Generation ### Fixes Fixed an issue where the Application/Bot user sees the URL of their Aisera tenant application instead of the Knowledge Base Article URL.