These release notes provide a list of changes to AI features in Pendo as they occur, so you can learn what’s new and reference the relevant documentation.
June 2026
View volume trends for use cases and issues in Agent Analytics
You can now see how your tracked and emergent use cases and issues are changing over time directly in Agent Analytics. Tracked use cases and issues now include a stacked bar chart above the existing table, showing conversation volume for each item across daily, weekly, or monthly time ranges. Emergent use cases and emergent issues now display as a treemap chart with trend indicators that show whether volume is growing compared to the prior equivalent period. Selecting any chart element opens a panel with the relevant conversations so you can investigate without leaving the current view. For more information, see Analyze and track use cases for your AI agents and Identify and track issues with your AI agents.
Ask Pendo questions directly in Slack (beta)
May 2026
Enable write tools for the Pendo MCP server
You can now enable write tools for the Pendo MCP server, allowing connected AI clients to take actions in Pendo in addition to reading data. The first available write tool creates Listen feedback items, letting you capture feedback directly from AI clients like Claude or Cursor without switching to Pendo. For more information, see Pendo MCP tools for AI agents.
Leo supports funnel analysis
You can now ask Leo about conversion and drop-off rates through a funnel. You can also filter results by applying segments within the same conversation. This means you no longer need to create a report to analyze funnel behavior. For more information, see Analyze product usage with Leo.
Auto-fill business context in settings
Subscription admins can now generate business context using AI. Business context informs AI features such as Leo and Listen Explore about your company, so you can receive tailored results based on your business, product offerings, and competitors. In subscription settings, provide a business website and select Generate business context. Pendo uses Tavily, a third-party AI service, to generate the context based on the URL provided. You can make changes to the AI-generated output before saving. For more information, see Subscription settings.
Compare AI agent versions
You can now run experiments on your AI agent conversations to measure the impact of changes to your agent configuration. This lets you compare conversation metrics side-by-side between control and experiment groups to validate model changes, tool updates, and other agent modifications with data-driven insights, helping you iterate confidently and speed up the feedback loop from change to outcome. For more information, see Compare AI agent versions with experiments.
Ask Leo about your Agent Analytics data
You can now use Leo, Pendo's AI assistant, to have conversational interactions with your Agent Analytics data. Ask questions about patterns, themes, and drivers behind issues and use cases to quickly extract insights from your customer conversations. This helps you understand customer needs and prioritize work without manually analyzing charts and filters. For more information, see Analyze and track use cases for your AI agents and Identify and track issues with your AI agents.
Send Agent Analytics digests to Slack
You can now send scheduled Agent Analytics digests to a Slack channel, so your team can monitor AI agent performance without leaving Slack. Digests include key metrics, visitor adoption, and emergent issues for a selected agent and segment on a recurring schedule. For more information, see Send Agent Analytics digests to Slack.
Track issues in Agent Analytics
You can now save and monitor specific problems in your AI agent conversations as tracked issues in Agent Analytics. Pendo auto-generates detection rules from your description, which you can edit or add to. This helps you measure ongoing impact and prioritize improvements over time. For more information, see Identify and track issues with your AI agents.
Use skills in Leo conversations
You can now use skills in Leo to follow guided workflows for consistent results. Skills are pre-defined prompts that help you accomplish a specific task, such as preparing for a customer call, reviewing an account, or analyzing product health. To use a skill, select a tile on the Leo homepage, enter the corresponding slash command, or ask a question in natural language. For more information, see Analyze product usage with Leo.
Signals in open beta
If you're using Leo, the new homepage now surfaces Signals alongside Ask Leo and skills in a single, consolidated view. Signals are generated automatically from your product usage data to highlight notable changes in retention and adoption, and you can investigate any signal directly with Leo to understand what changed and why. For more information, see Signals on your homepage and Analyze product usage with Leo.
April 2026
Issue types in Agent Analytics
You can now see a breakdown of issues by type in Agent Analytics. Issues are categorized as rage prompts, unsupported requests, or errors, each shown as a percentage of total conversations. This helps you identify where your agent is struggling and prioritize improvements. For more information, see Identify issues with your AI agents.
Leo can answer questions about using Pendo
You can now ask Leo questions about how to use Pendo and it'll refer to Pendo Help Center articles in its answers. Ask Leo how to do something in Pendo and it returns a summary answer with links to the relevant Help Center articles. For more information, see Analyze your product data with Leo.
Provide proactive support with Intercom's Fin AI Agent and Pendo MCP server
You can now integrate Pendo with Intercom's Fin AI Agent to provide real-time, in-app support when your users experience friction. Send frustration events (such as rage clicks, error clicks, or U-turns) from Pendo to Intercom using webhooks, which trigger an Intercom outbound message that opens a proactive Fin conversation to offer assistance. For more information, see Connect behavioral data to the Fin AI Agent.
Fetch developer logs through the Pendo MCP server
You can now use the Pendo MCP server to retrieve developer logs, including console output and network request and response details, from a specific replay. This allows you to use your own AI tools to investigate technical issues with the full context of what happened in the browser during a session, without leaving your AI environment. This requires Session Replay and the Pendo MCP server to be turned on. For more information, see Pendo MCP tools for AI agents.
Leo is available to subscriptions with a BAA
You can now use Leo if your Pendo subscription has a Business Associate Agreement (BAA). Leo uses OpenAI as one of its underlying AI models, which now supports BAAs. This allows organizations in healthcare and other regulated industries to use Leo. A subscription admin needs to enable Leo in the subscription settings to make it available to users. For more information, see Artificial intelligence (AI) at Pendo.
Simplified setup flow for Agent Analytics
You can now add an AI agent in Pendo using a setup flow that automatically routes you to the right capture configuration based on your app type. Web and mobile apps go directly into full conversation capture setup, while browser extension apps offer a choice between prompts-only and full conversation capture. New agent context fields let you describe your agent's role, purpose, and intended users, providing Pendo with more information to surface relevant insights. For more information, see Add and configure AI agents in Pendo.
Leo is available in the US-1 environment
You can now use Leo in Pendo subscriptions hosted on US-1 data centers (app.pendo.io). Previously, Leo was only available for US and EU data centers. Enable Leo through your subscription settings to get started. For more information, see Analyze product usage in agent mode.
Pendo MCP server now generally available 
The external Pendo Model Context Protocol (MCP) server is now generally available, letting you bring live Pendo product and customer data into MCP‑supported AI clients such as Claude, ChatGPT, Cursor, Gemini CLI, and Windsurf. You can use the MCP server to access visitor and account metadata, query application analytics and user behavior, and search for objects like Pages, Features, Track Events, guides, and AI agent entities directly from where you already work. For more information, see Connect to the Pendo MCP server.
March 2026
Agent mode is now Leo
You can now use Leo, previously known as Agent mode, to ask questions about your product usage data in a conversational interface. Leo covers adoption, retention, user sentiment, and account behavior across Pages, Features, Track Events, feedback, and related metadata. Select Ask Leo from Pages, Features, Track Events, Accounts, or Visitors to get started. Leo offers a faster way to explore your data and answer product questions without building reports. For more information, see Analyze product usage with Leo.
Agent mode is available in the EU environment
You can now use Agent mode in Pendo subscriptions hosted on EU data centers (app.eu.pendo.io). Previously, Agent mode was only available for US-hosted subscriptions. Enable Agent mode through your subscription settings to get started. For more information, see Analyze product usage in agent mode.
Reference Product Areas in agent mode with @mentions
You can now use the @ mention syntax to reference Product Areas when asking questions in agent mode. This helps you scope your questions to specific parts of your product without manually filtering. For more information, see Analyze product usage in agent mode.
New MCP tools for AI agents and feedback
You can now use four new tools with the Pendo MCP server: list_ai_agents, list_use_cases, list_ai_agent_issues, and get_ideas. These tools let external AI clients query Agent Analytics data and Listen ideas directly. For more information, see Pendo MCP tools for AI agents.
Expanded access to Agent Analytics
You can now access Agent Analytics without a subscription admin role. Non-admin users can view the agent overview, use cases, and existing reports, but viewing visitor prompts and conversations still requires the subscription admin role or the AI Agent Admin role for that app. For more information, see Overview of Agent Analytics.
Filter Agent Analytics by conversation metadata
You can now filter Agent Analytics data by conversation-level metadata, including user reactions, whether a suggested prompt was used, and whether a file was uploaded. Filters apply across each tab and in dashboard widgets. For more information, see Analyze interactions with AI agents.
Refine tracked use cases with example prompts
You can now add up to 10 example prompts to each tracked use case in Agent Analytics. Example prompts help define which requests should be grouped together, improving how accurately related prompts and conversations are matched over time. For more information, see Analyze and track use cases in Agent Analytics.
New overview charts for AI agents
Agent Analytics now includes additional overview charts for agents that send full conversation data through the Conversations API. These charts help you understand how visitors react to responses, how frequently suggested prompts are used, and which models are used the most. For more information, see Analyze interactions with AI agents.
New tab for AI settings
You can now find all AI settings in a dedicated tab through Subscription settings > AI access. AI settings are grouped by product area for Agent mode, Pendo MCP, Analytics, Guides, Listen, Session Replay, and Sentiment. For each AI feature, you can see which underlying AI models are being used. This provides additional clarity to subscription admins who manage these settings. For more information, see Subscription settings.
Use an AI skill for quicker conversation setup
You can now use the Pendo setup-agent-analytics skill with AI coding assistants like Claude Code or Cursor to automate Conversations API implementation for Agent Analytics. This reduces the time and effort needed to start capturing full conversation data in Agent Analytics. For more information, see Automate Conversations API implementation with an AI skill.
Send server-side conversation events for mobile and non-web apps
You can now send AI agent conversation events to Pendo from your backend using the server-side Conversations API in Agent Analytics. This lets you capture full conversations from mobile apps, browser extensions, and other environments where the Pendo Web SDK isn't available. For more information, see Capture full conversations with the Conversations API.
February 2026
Preview conversations mode in Agent Analytics
You can now preview the Issues and Conversations tabs for prompt-only agents. This preview displays sample data to help you understand the value of capturing full conversations before installing the Conversations API. For more information, see Preview conversations mode.
Capture AI agent issue details for external tickets
You can now copy structured issue details from Agent Analytics to include in an external ticketing system, such as Jira or Linear. This makes it easier to share relevant context, supporting conversation examples, and key metadata with engineering or support teams, helping streamline collaboration and improve issue resolution. For more information, see Identify issues with your AI assistants.
Adjust detail of emergent use cases in Agent Analytics
You can now adjust the level of detail for emergent use cases in Agent Analytics using a new specificity slider. This control lets you shift between broader themes and more granular clusters, helping you tailor results to your investigative goals. The selected level of detail can also be saved to an Agent Analytics report. For more information, see Analyze and track use cases in Agent Analytics.
New conversation-level metrics in Agent Analytics
Agent Analytics now includes two additional key metrics for conversations: Issue rate and Average prompts per conversation. Issue rate shows the percentage of conversations that encountered at least one issue, with issues deduplicated at the conversation level, while Average prompts per conversation shows how many prompts occur, on average, within a single conversation. These metrics are available only for agents capturing full conversations and provide quick insight into conversation quality and interaction effort. For more information, see Analyze interactions with AI agents.
Share Agent Analytics conversations and view response data
You can now share specific prompts or agent responses when viewing conversations in Agent Analytics. Shared links load the relevant agent and date range in Pendo and open directly to the selected prompt or response, making it easier to review conversations with teammates or return to a specific interaction later. Conversations also now surface additional response data within the conversation panel, including which tools were called (if any) and which model was used to generate a response.
Agent mode now in open beta
You can now use agent mode to ask natural-language questions about your product usage data instead of building reports manually, helping you explore adoption, retention, sentiment, and account behavior across Pages, Features, Track Events, feedback, and related metadata in a conversational interface. You can also use a guided account health prompt to generate a shareable summary before a customer call, ask about trends in qualitative feedback, or draft a guide based on what you learn. For more information, see Analyze product usage in agent mode.
Query customer feedback with Listen tools in MCP
You can now use Listen tools with the external Pendo MCP server to query customer feedback. These tools let you retrieve raw feedback items, view extracted feedback insights with summaries and supporting quotes, and generate AI-clustered topics that highlight common themes in filtered feedback. For more information, see MCP query tools.
January 2026
New conversation-related columns for Agent Analytics
The Conversations table in Agent Analytics now includes additional columns that provide more context about each interaction. New columns indicate whether a conversation contains rage prompts, includes a file upload, uses a suggested prompt, calls tools, or uses a specific model. These additions help teams quickly identify conversations that may require deeper review or indicate more complex agent behavior. For more information, see Analyze interactions with AI agents.
December 2025
Use semantic search with the MCP server
You can now use semantic search with the Pendo MCP server to match Pages, Features, Track Events, and Guides based on meaning rather than exact names when querying your data from external AI clients. This can help AI-powered workflows return relevant results even when object names aren’t known precisely. Semantic search is also available in agent mode if you’re included in the closed beta.
Agent Analytics now generally available
Agent Analytics helps you understand how visitors interact with your AI-powered agents by capturing prompts or conversations, surfacing emergent and tracked use cases and issues, and highlighting signals like rage prompts and retention. For more information, see our Agent Analytics documentation.
Identify issues in AI agent conversations (beta)
Agent Analytics now surfaces emergent issues to help you quickly identify potential problems in AI agent conversations. Issues are detected automatically from conversation content and grouped by theme and severity, making it easier to spot patterns that may indicate confusion, failure to complete tasks, or user frustration. For more information, see Identify issues with your AI assistants.