Memory

Memory enables agents to remember information from previous interactions, allowing them to maintain context across conversations. Without memory, each agent interaction would be isolated, requiring users to repeat information in every message. With memory, agents can reference past conversations, learn from feedback, and adapt to user preferences over time. Memory stores conversation history associated with session_id and user_id. When you make a request with the same identifiers, the agent automatically includes relevant conversation history in the context, enabling it to reference previous messages and maintain continuity. Session vs User IDs:

session_id - Isolates conversation threads. Each session maintains separate context.
user_id - Groups conversations by user. Useful for maintaining user-specific context across sessions.

You cannot define session_id alone. Use user_id for long-term memory (persists across sessions) or both session_id and user_id together for short-term memory (isolated to a specific session).

Supported Memory Backends

Hypertic supports multiple memory backends. Click on a backend to see setup and usage examples:

MongoDB

View guide >

PostgreSQL

View guide >

Redis Cache

View guide >

InMemory

Use InMemory for testing and development. Data is stored in RAM and lost when the process ends:

import os
from dotenv import load_dotenv
from hypertic.memory import InMemory
from hypertic.agents import Agent
from hypertic.models import XAI

load_dotenv()

def main():
    memory = InMemory()
agent = Agent(
        model=XAI(
            model="grok-3",
            max_tokens=4096,
        ),
        memory=memory
    )
    
    user_id = "user_123" 
    session_id = "session_001"
    
    print("Chat with the agent (type 'quit' to exit):\n")
    
    while True:
        user_input = input("You: ").strip()
        if user_input.lower() in ['quit', 'exit', 'q']:
            break
        if not user_input:
            continue
        try:
            response = agent.run(user_input, session_id=session_id, user_id=user_id)
            print(f"Agent: {response.content}\n")
        except Exception as e:
            print(f"Error: {e}\n")


if __name__ == "__main__":
    main()

For production applications, use persistent storage like PostgreSQL or MongoDB. See the individual backend pages for setup instructions.

Guardrails Model Context Protocol (MCP)

Get started

Agent

Workflow

Supported Memory Backends

MongoDB

PostgreSQL

Redis Cache

InMemory

​Supported Memory Backends

MongoDB

PostgreSQL

Redis Cache

​InMemory

Supported Memory Backends

InMemory