Source record · tier 2 current vendor documentation
Webex AI Agent Studio Administration guide
- Publisher
- Cisco Systems, Inc. (Webex Help Center)
- URL
- https://help.webex.com/en-us/article/ncs9r37/Webex-AI-Agent-Studio-Administration-guide
- Published
- 2026-09-03
- Updated
- unknown
- Accessed
- 2026-09-04
- HTTP status
- 200
- License
- Cisco proprietary documentation, all rights reserved; no-redistribution; short excerpts and locators only
Source notes citing this source
- On digital channels, the Webex AI Agent transparency note is shown once at the start of the AI-agent session with the first AI-generated response. in context
- Outside European regions, Webex AI Agent administrators choose whether to enable the AI transparency disclosure and can configure its message; Cisco recommends enabling it as a best practice. in context
- The 90-day transcript purge documented for Webex AI Agent Studio is not stated to apply to AI Receptionist for Webex Calling, because that guide documents Webex Contact Center AI agents and never names AI Receptionist. inferred in context
- In Webex AI Agent Studio no user has access to transcripts by default; a full administrator must explicitly grant decrypt access to individual users, and locked sessions require that permission to view. in context
- The Webex AI Agent Studio management guide states that transcripts are stored in Webex AI Agent Studio for 90 days, after which they are automatically purged. in context
- In Webex AI Agent Studio the Agent handover action, which lets the AI agent escalate the conversation to a human agent, is enabled by default and can be turned off with a toggle. in context
- For the Webex AI Pro-Europe 1.0 engine, the default English voice is en-GB-Grace. in context
- For the Webex AI Pro 1.0 engine, the default English voice is en-US-Jennifer. in context
- For the Webex AI Pro-US 1.0 engine, the default English voice is en-US-Jess. in context
- The AI engine chosen for an autonomous agent determines its conversation settings, including language, voice, DTMF, and delays and interruptions. in context
- With the Webex AI Pro-US 1.0 or Webex AI Pro-Europe 1.0 engine, voice conversations can be previewed in English only. in context
- For the Webex AI Pro 1.0 engine, the default language is English and the administrator can choose another language from a drop-down list. in context
- English is the only supported language for Webex AI Pro-US 1.0, with en-US as the default locale, and for Webex AI Pro-Europe 1.0, with en-GB as the default locale. in context
- An autonomous agent's speaking rate accepts values from 0.7 to 1.2, and the default of 1.0 is normal speed. in context
- The voice is chosen from a Select voice drop-down, and the voices offered depend on the chosen AI engine and the configured language. in context
- Webex AI Agent is licensed as an add-on to the Cisco Collaboration Flex 3.0 Contact Center license, not as a Webex Calling user license. in context
- For enterprise contact centre deployments the Webex AI Agent add-on is purchased in Cisco Commerce Workspace, with ordering detail in the Collaboration Flex Plan Contact Center Ordering Guide. in context
- To use Webex AI Agent with Webex Contact Center, a customer buys the AI Agent add-on to the Flex 3.0 Contact Center licence. in context
- If an organisation has no active contact center trial or subscription, a partner can set up a contact center trial that includes the Webex AI Agent feature. in context
- As part of PCI compliance, autonomous AI agent fulfilment does not currently support flows involving debit or credit cards, and card handling must be done through a third-party integration instead. in context
- For autonomous agents, Agent handover is a built-in action that is enabled by default and can be toggled off, rather than a set of rule-based triggers. in context
- An autonomous AI agent is configured from free-text instructions plus an optional knowledge base and a set of declared actions with JSON-schema-typed input entities, and the model decides at run time which action to invoke. in context
- Cisco states that it opts in to the Content filtering feature provided by Azure OpenAI to prioritise content safety, while acknowledging that generative AI features are prone to errors. in context
- For Contact Center Enterprise and Packaged CCE, the Webex AI Agent add-on is ordered in Cisco Commerce Workspace against the Collaboration Flex 3.0 Contact Center offering and requires a hybrid organisation setup that integrates the cloud and on-premises environments. in context
- For Contact Center Enterprise and Packaged CCE, Webex AI Agent voice support requires Cisco Unified CVP, Cisco VVB and Cloud Connect to be on release 15.0(1) ES202511 or later. in context
- For cloud Webex Contact Center, access to AI Agent Studio requires purchasing the AI Agent add-on for the Flex 3.0 Contact Center licence. in context
- Beyond licensing, the Webex AI Agent option in Control Hub is gated behind a Cisco-enabled feature flag; if the option is not visible the organisation must contact Cisco Support to have it enabled. in context
- Cisco instructs administrators not to upload Payment Card Industry data, personally identifiable information or protected health information to the AI Agent Studio knowledge base, and makes the customer responsible for upload compliance. in context
- Cisco tells customers with EU-facing deployments to review the default transparency message and confirm the final wording meets their branding, legal and compliance requirements before 2 August 2026. in context
- AI Agent Studio's runtime guardrails are described only as preventing unethical and harmful responses; no guardrail configuration object, screen or setting is documented anywhere in the administration guide. inferred in context
- An AI Agent Studio knowledge base is capped at 2 GB of total storage and 100 files, with an individual file size limit of 10 MB and a lower limit of 2 MB for .txt files. in context
- An autonomous AI agent can be configured with a maximum of 10 actions, counting fulfilment, MCP client and custom transfer actions together. disputed in context
- An organisation can create up to 100 AI agents in AI Agent Studio, counting scripted and autonomous agents together. in context
- In its MCP client action limitations, the same guide states that the maximum number of actions per agent is defined by the organisation's configuration and directs the reader to their administrator. disputed in context
- Cisco states that Microsoft does not use customer content to improve Azure OpenAI models and does not store or retain Cisco customer data in Azure infrastructure. in context
- Cisco states that customer input data is not retained after inference and that the third-party model provider, Microsoft, does not access, monitor or store Cisco customer data. in context
- Out-of-the-box reports for AI agents are not available in Analyzer; an administrator must create a flow global variable and import 2 supplied custom visualisation templates to obtain AI agent reporting. in context
- Cisco applies automated pattern detection for PCI data in AI Agent voice recordings and limits recording visibility when it fires, but states explicitly that the detection may not identify every instance. in context
- Advanced debug logs associated with a scripted agent transaction are typically retained for 180 days. in context
- For scripted agents, handover to a human can be triggered by 4 configurable conditions: consecutive partial matches, consecutive fallback messages, consecutive repeated agent responses, or reaching the slot-filling retry limit. in context
- A scripted AI agent's knowledge base is a corpus of questions and answers that the administrator supplies, and the natural language understanding model must be retrained after the corpus is edited. in context
- An autonomous agent selects 1 of 3 named AI engines: Webex AI Pro 1.0, the multi-language default; Webex AI Pro-US 1.0, English only and available to US customers only; and Webex AI Pro-Europe 1.0, English only and available to Europe customers only. in context
- No user has access to AI agent transcripts by default; a full administrator must explicitly grant that access, including to other full administrators. in context
- AI Agent conversation transcripts are stored for 90 days and then automatically purged, stated identically for autonomous and scripted agent sessions. in context
- AI transparency, the disclosure telling a customer they are speaking to an AI, is enabled by default in European regions and optional elsewhere; disabling it requires the administrator to acknowledge the risk and record audit comments. in context
- Webex AI Agent Studio offers exactly 2 AI agent types, scripted and autonomous. in context
- Autonomous voice conversation timing is configurable: end-of-speech sensitivity 500 to 2000 milliseconds with a 500 millisecond default, fulfilment timeout 10 to 30 seconds defaulting to 30, caller turn timeout 750 to 3000 milliseconds defaulting to 1500, no-input timeout 10 to 30 seconds defaulting to 10, DTMF inter-digit timeout 2 to 10 seconds defaulting to 5, DTMF maximum input length 8 to 32 characters defaulting to 16, and speaking rate 0.7 to 1.2 defaulting to 1.0. in context
- For autonomous AI agents in Webex AI Agent Studio, the DTMF Digit Timeout between two digits defaults to 5 seconds and can be 2 to 10 seconds. in context
- In Webex AI Agent Studio DTMF settings, Max length for caller DTMF input defaults to 16 characters and can be 8 to 32. in context
- In Webex AI Agent Studio DTMF settings, the termination character that ends caller DTMF input can be # or *. in context
- Cisco states that Model Context Protocol actions are read-only once created and that changing an action's settings requires creating a new action. in context
- Cisco states that with a Model Context Protocol client configured an agent can automatically invoke third-party tools during live conversations, retrieve real-time data from external systems, and perform actions without human intervention. in context
- Model Context Protocol client actions appear only under the autonomous agent branch of the AI Agent Studio guide and nowhere in the scripted agent workflow or its contents. inferred in context
- Using AI Agent Studio with Contact Center Enterprise requires Unified Customer Voice Portal, Cisco Virtualized Voice Browser and Cloud Connect on release 15.0(1) ES202511 or later. in context
- Access to the AI Agent Studio change logs requires an administrator or artificial intelligence agent developer role, or a custom role carrying the audit log permission. in context
- The Model Context Protocol client capability lets autonomous artificial intelligence agents connect directly to third-party tools and services without custom-built interface integrations. in context
- The AI Agent Studio guide sends the reader to sections named Onboard your Agent App and Provision on Control Hub, and no page retrieved carries either title; the nearest are pages titled Onboard Your MCP Server and a link labelled Provision Agentic Apps in Control Hub. in context
- Cisco states that group-level access control is not supported for Model Context Protocol tools and that tool access is managed at the organisation level only. in context
- No source read states what data leaves Webex during a Model Context Protocol tool call, how long a third party retains it, or any residency logging or audit commitment specific to those calls. inferred in context
- The AI Agent Studio guide documents no timeout, retry count or agent fallback behaviour for a failed Model Context Protocol call, while documenting several timeouts for other action types. inferred in context
- Cisco states that autonomous agent fulfilment does not support debit or credit card handling for payment card compliance and does not say whether the same restriction applies to Model Context Protocol client actions. in context
- Cisco states that the maximum number of actions an agent can support is defined by the organisation's configuration and gives no number, telling the reader to contact their administrator. disputed in context
- For each Model Context Protocol action Cisco exposes the tool name description and server details, the input parameters sent, the output returned, an execution state of success or failure, latency in milliseconds, the server name and identifier, and a transaction identifier. in context
- Cisco's statement that it does not retain customer input after inference and that Microsoft does not access monitor or store Cisco customer data concerns the underlying language model provider and not the third-party servers an agent calls. in context
- The AI Agent Studio guide does not state which contact centre products support Model Context Protocol client actions specifically; its availability table lists agent types against 5 solutions without a per-cell grid. inferred in context
- Nothing in the AI Agent Studio guide restricts what an underlying Model Context Protocol tool may do; the read-only statement concerns the action's configuration rather than the operation it performs. inferred in context
- Modifying a Model Context Protocol action's automatically populated name, description or input entities requires updating the server definition at the registration or provisioning stage, and Cisco states those changes synchronise automatically to AI Agent Studio. in context
- Cisco states that Webex AI Agent Studio supports only 3 authentication types for a Model Context Protocol client action: OAuth2 client credentials, application key, and custom header based authentication. disputed in context
- A different section of the same AI Agent Studio guide states a flat maximum of 10 actions for an artificial intelligence agent, covering Model Context Protocol client actions, fulfilment actions and custom transfer actions together. disputed in context
- Enabling a Model Context Protocol action is a 3-stage pipeline: register the application on the Webex developer portal, authorise it in Control Hub for the organisation, then configure the action in AI Agent Studio. in context
- Adding a Model Context Protocol tool automatically populates the action name from the tool name, its description, and its input entities including required fields and data types, from the server definition. in context
- AI Agent Studio is reached from either Control Hub or Webex Connect, and voice preview for agents is available only through the Control Hub path. in context
- Artificial intelligence agent support on Unified CCX is stated to be an alpha release. in context
- For the Webex AI Pro-Europe 1.0 engine, English is the only language supported by default and en-GB is the default locale. Webex AI Pro 1.0 supports multiple languages. in context
- For the Webex AI Pro-US 1.0 engine, English is the only language supported by default and en-US is the default locale. in context
- AI Agent Studio transfer actions have a toggle: when on, the AI agent tells the customer about the transfer before executing it; when off, it performs a silent transfer. in context
- In Webex AI Agent Studio the Agent handover action is enabled by default, allowing the AI agent to escalate the conversation to a human agent. in context
- Webex AI Agent Studio keeps session transcripts for 90 days and then purges them automatically. in context
Cite this source record
APA
WarmTransfer. (2026, September 3). Webex AI Agent Studio Administration guide. WarmTransfer. https://warmtransfer.net/knowledge/sources/cisco-help-ncs9r37-wxcc-ai-agent-studio-admin
BibTeX
@misc{warmtransfer-cisco-help-ncs9r37-wxcc-ai-agent-studio-admin,
title = {Webex AI Agent Studio Administration guide},
author = {{WarmTransfer}},
year = {2026},
url = {https://warmtransfer.net/knowledge/sources/cisco-help-ncs9r37-wxcc-ai-agent-studio-admin},
note = {Cisco Systems, Inc. (Webex Help Center), accessed 2026-09-04}
}