Name: Letta AI
Uploaded: 2025-02-15T07:35:54.430Z
Duration: 2 min 33 s
Description: An open-source platform for building stateful AI agents with long-term memory and advanced reasoning.

letta-ai

An open-source platform for building stateful AI agents with long-term memory and advanced reasoning.

Letta AI

Hey there. I'm Charles, the CEO of Letta and creator of memgpt. Letta makes it easy to build agents with persistent memory and advanced reasoning using any model you want, both open and closed. You can build agents with Letta using our agent development environment or ad. 

All right, so first I'm going to create an agent with the ad. The server is configured with open AI, but you can use any other model provider like anthropic or even open source models like llama 3. You can customize the starter data for the agent's memory. This defines the agent's persona. You can also customize the personalization information for the user. All right, now we have a running agent, so let's message it. You can see that the agent is streaming inner thoughts. This is similar to Open AI's new models like 01 that can think for an extended period of time before they act, which helps the agent have.

More advanced reasoning capabilities. On the left side we can see the memories of the agent both the core memories which are in context and archival which are out of context as well as the tools that the agent has access to. The tools include both memory tools and the default lettera tools you.

Can always add your own custom tools to agents. Let agents are special because they can learn and adapt over time thanks to their memories. Let's message the agent with a correction about our name and see what happens. We can see here how the agent used the memory tool to update its own.

Memory like the name implies the ad is a development environment. So the agent you create here is intended to be a backend service to your application. You can use these toggles to see what is exposed to your end user. You can also control the starter message that triggers an agent to run since.

This agent lives on a server. You can also message the agent from outside the ad using the agent's unique ID. So let's go to the terminal and send a message to the agent. 

So we see that the message also showed up on the ad since the ad is communicating with the same Leta agent server as your terminal we're excited to see what products you build using Leta. Check out our developer documentation and our other video tutorials to start building for.

Note, that there may be slight differences in text and video. Watch video for the original content.

Letta AI Video