Teams across product, support, onboarding, and success are all deploying agents, with AI investment on track to reach 3.3 trillion dollars by 2027. The bet companies are making on agents is enormous.
But there’s a mounting problem: companies still aren’t seeing ROI on their investments. In fact, 34% of consumers say AI support has actually made things harder, not easier. The challenge is that agents don’t really know the user they are trying to help. They don’t know what the user was doing before interacting with the agent or that user’s history in the application. That context is critical for an agent to deliver truly personalized and helpful support.
This is the problem we’re solving with the Pendo Agent Toolkit, a new tool that connects the AI agents you build and buy, like Fin or Decagon, to real-time product context. With this data, in-app assistants and agents can proactively guide, support, and act on behalf of users.
Agents have plenty of data, but that’s not enough
Agents today aren't completely in the dark. They're working from a subset of data, like telemetry, firmographics, revenue, CRM records, and marketing behavior. They’re also trained on extensive knowledge base articles, in depth instructions, and other generic content.
However, none of that tells the agent what a user is doing in the application at a given moment. It can’t see if the user ever finished account setup, or that they've only touched 2% of the features they're paying for, or that they rage-clicked 17 times before the conversation started.
That real-time product context is what agents have been missing, and it's the input that determines what support is needed, and ultimately, whether a response delivers value or not.
Pendo has spent 13 years capturing product usage data on behalf of customers, and helping you take action based on insights from your data.. Now, with Agent Toolkit, you can connect that data to your agents, so they can take smarter actions autonomously. It works with agents you bought like Fin, Decagon, or Sierra, or those you’ve built in-house and embedded in your application.
Two mechanisms: MCP for context, webhooks for real-time triggers
Agent Toolkit runs on two mechanisms:
- MCP passes context to your agent, so it understands the user's journey. This includes login activity, feature usage, adoption metrics, retention insights, and engagement scores. It also gives the agent access to the guides you've already built in Pendo, so it can surface them when they're relevant.
- Next, webhooks fire when something happens, like a rage click or a dropped workflow, so the agent can step in on its own.
Together, this forms the context and action layer that agents need to be effective.
Five things Agent Toolkit lets your agent do today
- Real-time session context: Through Pendo’s MCP server, agents get access to a user’s in-app history, like where they were before getting stuck, what features they use the most, and which features they’ve never touched. This is essential to understanding where each user is in their journey, and what they need help with.
- Proactive frustration detection: Real-time triggers alert agents when something goes wrong in-app. Pendo sends frustration signals, like rage clicks and U-turns, to agents in the moment so it can proactively offer help, before the user goes looking for it.
- Personalized and tailored guidance: Agents pick the right guide for each user's situation, whether that's a walkthrough, a tooltip, or anything else you've published in Pendo. No complex segmentation to maintain, and no guidance that only shows up if the user happens to land on the right page.
- Contextual workflow nudges: When a user starts a task and doesn't finish it, agents get notified. They can reach out with a specific prompt to bring them back to where they left off, rather than waiting for a user to rediscover the flow on their own.
- Feedback: When a user shares something negative mid-conversation, agents capture it and send it back to Pendo as feedback. The team that owns that part of the product sees it without anyone having to forward a transcript.
Built for support and product teams
Customer support teams
Support teams deploying agents inside their application, like Fin, Decagon, or Sierra, need a way to make them smarter and more personalized than their knowledge base. Agent Toolkit gives support agents the user's in-app history and the moment of frustration, so it can resolve the issue on first contact instead of sending in a ticket.
Product
With Pendo, product teams can turn their agent into an onboarding and adoption channel. The agent surfaces guides at the right moment, pulls users back into flows, and points them toward features they're paying for but haven't found. You can use the same guides you've already built, and you control what the agent can show.
Introducing Pendo for Agents: when product context meets observability
Agent Toolkit is one half of something bigger: Pendo for Agents, our answer to the belief that agents only get smarter when they know what's actually happening in your product.
The other half is Agent Analytics. Once your agents have the context to act in the moment, you'll have the data to prove it's driving real outcomes, like fewer tickets, faster resolutions, and happier users.
If you’re already using Pendo, Agent Toolkit is available for free in open beta. If you’re not, talk to an expert to get started.
