Skip to content

memory.md

Latest commit

 

History

History
132 lines (100 loc) · 5.68 KB

memory.md

File metadata and controls

132 lines (100 loc) · 5.68 KB
title Step 4: Memory & Persistence
description Add context providers and persistent memory to your agent.
zone_pivot_groups programming-languages
author eavanvalkenburg
ms.topic tutorial
ms.author edvan
ms.date 02/09/2026
ms.service agent-framework

Step 4: Memory & Persistence

Add context to your agent so it can remember user preferences, past interactions, or external knowledge.

:::zone pivot="programming-language-csharp"

By default, agents will store chat history in an InMemoryChatHistoryProvider or in the underlying AI service, depending on what the underlying service requires.

The following agent uses OpenAI Chat Completion, which neither supports nor requires in-service chat history storage so therefore automatically creates and uses an InMemoryChatHistoryProvider.

using System;
using Azure.AI.Projects;
using Azure.Identity;
using Microsoft.Agents.AI;

var endpoint = Environment.GetEnvironmentVariable("AZURE_OPENAI_ENDPOINT")
    ?? throw new InvalidOperationException("Set AZURE_OPENAI_ENDPOINT");
var deploymentName = Environment.GetEnvironmentVariable("AZURE_OPENAI_DEPLOYMENT_NAME") ?? "gpt-4o-mini";

AIAgent agent = new AIProjectClient(new Uri(endpoint), new DefaultAzureCredential())
    .AsAIAgent(
        model: deploymentName,
        instructions: "You are a friendly assistant. Keep your answers brief.",
        name: "MemoryAgent");

Warning

DefaultAzureCredential is convenient for development but requires careful consideration in production. In production, consider using a specific credential (e.g., ManagedIdentityCredential) to avoid latency issues, unintended credential probing, and potential security risks from fallback mechanisms.

To use a custom ChatHistoryProvider you can pass one to the agent options:

using System;
using Azure.AI.Projects;
using Azure.Identity;
using Microsoft.Agents.AI;

var endpoint = Environment.GetEnvironmentVariable("AZURE_OPENAI_ENDPOINT")
    ?? throw new InvalidOperationException("Set AZURE_OPENAI_ENDPOINT");
var deploymentName = Environment.GetEnvironmentVariable("AZURE_OPENAI_DEPLOYMENT_NAME") ?? "gpt-4o-mini";

AIAgent agent = new AIProjectClient(new Uri(endpoint), new DefaultAzureCredential())
    .AsAIAgent(model: deploymentName, options: new ChatClientAgentOptions()
    {
        ChatOptions = new() { Instructions = "You are a helpful assistant." },
        ChatHistoryProvider = new CustomChatHistoryProvider()
    });

Use a session to share context across runs:

AgentSession session = await agent.CreateSessionAsync();

Console.WriteLine(await agent.RunAsync("Hello! What's the square root of 9?", session));
Console.WriteLine(await agent.RunAsync("My name is Alice", session));
Console.WriteLine(await agent.RunAsync("What is my name?", session));

Tip

See here for a full runnable sample application.

:::zone-end

:::zone pivot="programming-language-python"

Define a context provider that stores user info in session state and injects personalization instructions:

:::code language="python" source="~/../agent-framework-code/python/samples/01-get-started/04_memory.py" id="context_provider" highlight="4,15-20,39":::

Create an agent with the context provider:

:::code language="python" source="~/../agent-framework-code/python/samples/01-get-started/04_memory.py" id="create_agent" highlight="11":::

Run it — the agent now has access to the context:

:::code language="python" source="~/../agent-framework-code/python/samples/01-get-started/04_memory.py" id="run_with_memory" highlight="1,4,8,12,16":::

Tip

See the full sample for the complete runnable file.

Note

In Python, persistence/memory is handled by ContextProvider and HistoryProvider implementations. BaseContextProvider and BaseHistoryProvider remain as deprecated aliases, and InMemoryHistoryProvider is the built-in local, in-memory history provider. RawAgent may auto-add InMemoryHistoryProvider() in specific cases (for example, when using a session with no configured context providers and no service-side storage indicators), but this is not guaranteed in all scenarios. If you always want local persistence, add an InMemoryHistoryProvider explicitly. Also make sure only one history provider has load_messages=True, so you don't replay multiple stores into the same invocation.

You can also add an audit store by appending another history provider at the end of the list of context_providers with store_context_messages=True:

from agent_framework import InMemoryHistoryProvider
from agent_framework.mem0 import Mem0ContextProvider

memory_store = InMemoryHistoryProvider(load_messages=True) # add a history provider for persistence across sessions
agent_memory = Mem0ContextProvider("user-memory", api_key=..., agent_id="my-agent")  # add Mem0 provider for agent memory
audit_store = InMemoryHistoryProvider(
    "audit",
    load_messages=False,
    store_context_messages=True,  # include context added by other providers
)

agent = client.as_agent(
    name="MemoryAgent",
    instructions="You are a friendly assistant.",
    context_providers=[memory_store, agent_memory, audit_store],  # audit store last
)

:::zone-end

Next steps

[!div class="nextstepaction"] Step 5: Workflows

Go deeper: