Pendo Agent Analytics captures conversation data, including visitor prompts, agent responses, and session context, from your AI agents. Depending on how your application is built, you can send this data using one of several integration methods.
This article explains what's available and helps you choose the right path for your setup.
Available methods
Pendo supports the following methods for sending Agent Analytics data:
-
Client-side (Pendo Web SDK). Automatically captures conversation data from browser-based AI interfaces using the
trackAgent()method. Requires the Pendo Web SDK. - Server-side (HTTP API). Sends conversation data directly from your backend using standard HTTP requests. Required for mobile apps, backend services, and environments where the Pendo Web SDK isn't available.
- Python SDK (beta). Instruments Python-based AI agents to send conversation and trace data to Pendo. No Pendo Web SDK required.
- TypeScript SDK (beta). Instruments TypeScript-based AI agents to send conversation and trace data to Pendo. No Pendo Web SDK required.
You can also use the Pendo setup-agent-analytics skill with an AI coding assistant like Claude Code or Cursor to automatically add the correct instrumentation to your codebase. This skill supports both client-side and server-side methods.
Choose a method
If you want the most complete picture of your agents, including traces and execution details, use the Python SDK or TypeScript SDK (beta). These SDKs send both conversation events and trace data to Pendo.
Use this table to find the right integration path for your setup:
| Method | Best for | Requires Web SDK | Conversations | Use cases | Issues | Trace data |
|---|---|---|---|---|---|---|
| Pendo Web SDK | Extension apps with no code access (for example, Agentforce) | Yes ✅ |
No ❌ Prompts only; no agent response or user reaction |
Yes ✅ | No ❌ | No ❌ |
| Client-side (Conversations API) | Browser-based AI interfaces | Yes ✅ | Yes ✅ | Yes ✅ | Yes ✅ | No ❌ |
| Server-side (HTTP API) | Mobile apps, backend services, non-web environments (for example, CLI and voice agents) | No ❌ | Yes ✅ | Yes ✅ | Yes ✅ | No ❌ |
| Python SDK (beta) | Python-based AI agents; teams that need full visibility with traces | No ❌ | Yes ✅ | Yes ✅ | Yes ✅ | Yes ✅ |
| TypeScript SDK (beta) | TypeScript-based AI agents; teams that need full visibility with traces | No ❌ | Yes ✅ | Yes ✅ | Yes ✅ | Yes ✅ |
Tip: You can use the setup-agent-analytics skill to implement any of the client-side or server-side options above.
All of these methods send Agent Analytics data directly into the Pendo platform, so you can analyze AI agent activity alongside your existing product data in paths, funnels, and Session Replay.
Conversation events versus trace data
All integration methods can send conversation events, which are the prompts visitors submit and the responses your agent returns. This is the foundation of Agent Analytics and powers use case detection, issue identification, and conversation review.
Trace data provides more detailed information. When you send traces, Pendo captures the internal execution of your agent, including the steps it took, tools it called, models it used, and where errors occurred.
For a full explanation of what tracing is and how to view traces in Pendo, see Understand and review agent traces
Implementation steps
For step-by-step setup instructions, see the article for your method:
- Send conversation events using the client-side or server-side API
- Install the Python SDK to send conversation and trace data
- Install the TypeScript SDK to send conversation and trace data
- Automate Conversations API implementation with an AI skill
Shared requirements
Regardless of which method you use, you need:
- Full conversations activated for your agent in Pendo. For setup steps, see Add and configure AI agents in Pendo.
- The Agent ID for the agent you want to instrument. To find it, go to Product > Agent Analytics, then select the settings icon next to the agent name.
- Access to your product's codebase and a developer environment where you can deploy changes.
Processing and visibility
Conversation data is processed in hourly batches. After your code is live in production, new messages typically appear within 15 minutes after the start of the next batch. You can view data from Product > Agent Analytics by selecting the agent and opening the Conversations tab.
For more information on exploring Agent Analytics data, see Analyze interactions with AI agents.