Give your assistant a memory of each user with Mem0 (Python)

What you get

A short Python script that saves a conversation to Mem0's hosted platform under a user ID, then searches that user's memories with a plain-English question. Mem0 pulls the facts out of the messages (for example "Is a vegetarian" and "Allergic to nuts") and returns them with a relevance score, so your assistant can load them into its next prompt.

Prerequisites

A Mem0 Platform account.
A Mem0 API key in MEM0_API_KEY. Create it at https://app.mem0.ai/dashboard/settings?tab=api-keys&subtab=configuration
Python 3.10 or newer.
No second company's key is needed. The hosted platform does the memory extraction for you.

Steps

Install the SDK.

Shell

pip install mem0ai

Set your API key in the terminal you will run the script from.

Shell

export MEM0_API_KEY="your-api-key"

Create mem0_quickstart.py. The code is the official quickstart's three Python blocks joined into one file, with one change: the quickstart writes MemoryClient(api_key="your-api-key"), and here the client is created with no arguments so it reads MEM0_API_KEY from the environment (the SDK source falls back to that variable when no key is passed).

Python

from mem0 import MemoryClient

client = MemoryClient()

messages = [
    {"role": "user", "content": "I'm a vegetarian and allergic to nuts."},
    {"role": "assistant", "content": "Got it! I'll remember your dietary preferences."}
]
client.add(messages, user_id="user123")

results = client.search("What are my dietary restrictions?", filters={"user_id": "user123"})
print(results)

Test it

Run the script:

Shell

python mem0_quickstart.py

The quickstart shows the search returning a results list like this (shortened here to the main fields):

JSON

{
  "results": [
    {
      "id": "14e1b28a-2014-40ad-ac42-69c9ef42193d",
      "memory": "Allergic to nuts",
      "user_id": "user123",
      "categories": ["health"],
      "score": 0.87
    },
    {
      "id": "0f2c1b6e-9a3d-4b18-8f77-1c2d3e4f5a6b",
      "memory": "Is a vegetarian",
      "user_id": "user123",
      "categories": ["food_preferences"],
      "score": 0.81
    }
  ]
}

Your IDs, scores and categories will differ.

Not confirmed

The quickstart page does not say the Python client reads MEM0_API_KEY. That comes from the SDK source on GitHub (api_key or os.getenv("MEM0_API_KEY")), not from the docs page. If it does not work, pass the key as MemoryClient(api_key="...") as the quickstart shows.
The quickstart does not say whether a memory is searchable the instant add returns. If the first search comes back empty, run the script again after a few seconds.
Free tier limits and pricing were not checked.
The export line is from the quickstart's cURL tab and is for macOS and Linux shells. The docs do not show the Windows equivalent.
This script was not run as part of writing this recipe.

What you'll need

What you get

Prerequisites

Steps

Test it

Not confirmed