Meet Amplitude
AI Agents

Custom Agents let you set up an agent for a specific recurring job. Give it a name and instructions, point it at the workflow you want it to own, and it runs that job the same way every time — grounded in your Amplitude data and connected to the tools your team already uses. The purpose-built agents you may know from earlier this year carry forward as ready-to-use templates.

Agents work by synthesizing data from different parts of your Amplitude account: analytics, session replay, experimentation, guides and surveys, etc. So the more you're already using Amplitude to analyze user behavior, the more context your agents will have to learn from.

Amplitude agents are designed to work with a natural language interface, making them accessible to team members across all skill levels. You can interact with agents, access analysis, give feedback, and take action based on your comfort level.

Our agent system uses a variety of different models. The AI models we use to power agents currently are provided by OpenAI, Google, and Anthropic (served via OpenAI, Google Vertex, and AWS Bedrock), which use a large number of customized tools and subagents that do specific parts of the task. Most of it is homegrown backends for orchestration, memory, session storage etc. This has provided us with high leverage as we can fully customize the behavior for our agent system. In the near future, we plan to make cloud platform selection available in the AI Control Settings to enable or disable specific cloud providers OpenAI, Bedrock, and Vertex.

Agents are available as a core Amplitude platform—we want every customer to benefit from having AI teammates. We anticipate adding additional agentic capabilities across different plans, and there will be additional capabilities you can unlock with other products.

We do not train AI models on any customer data for Amplitude AI Agents.

Agents learn from a few different sources. They start with general best practices from the latest foundation models such as interaction patterns, accessibility standards, and design practices that might apply across different industries. From that foundation, we layer in Amplitude's expertise using system prompts that we've developed from working with thousands of teams.

Each agent builds up its own memory from each customer’s database about their specific product and what works for their users. This knowledge stays completely separate and secure for each customer. Over time, we expect agents will get smarter as their underlying models improve, as we refine our approach, and as they learn more about each particular business.