> ## Documentation Index
> Fetch the complete documentation index at: https://docs.mcp-use.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Memory management

> Control MCPAgent conversation history, stateless runs, and history resets.

`MCPAgent` keeps conversation history across runs by default. Disable memory for stateless requests, or clear history when a user starts a new task.

## Keep conversation memory

When `memoryEnabled` is `true`, the agent stores human, assistant, and tool messages after each run. This is the default.

```typescript theme={null}
import { MCPAgent } from "@mcp-use/agent";
import { MCPClient } from "@mcp-use/client";
import { ChatOpenAI } from "@langchain/openai";

const client = new MCPClient({
  mcpServers: {
    filesystem: {
      command: "npx",
      args: ["-y", "@modelcontextprotocol/server-filesystem", "./"],
    },
  },
});

const llm = new ChatOpenAI({ model: "gpt-4o" });

const agent = new MCPAgent({
  llm,
  client,
  memoryEnabled: true,
});

await agent.run({ prompt: "Hello, my name is Alice" });
const response = await agent.run({ prompt: "What's my name?" });

console.log(response);
await agent.close();
```

## Run without memory

When `memoryEnabled` is `false`, each run starts from the current system prompt and tool list only.

```typescript theme={null}
const agent = new MCPAgent({
  llm,
  client,
  memoryEnabled: false,
});

await agent.run({ prompt: "Hello, my name is Alice" });
const response = await agent.run({ prompt: "What's my name?" });

console.log(response);
```

Use stateless agents for request/response APIs, background jobs, and tests where earlier runs must not affect later behavior.

## Inspect history

`getConversationHistory()` returns a copy of the stored LangChain messages. Mutating the returned array does not mutate agent memory.

```typescript theme={null}
const history = agent.getConversationHistory();
console.log(`Current history has ${history.length} messages`);

for (const message of history) {
  console.log(message.constructor.name, message.content);
}
```

## Clear history

`clearConversationHistory()` removes stored conversation messages. If memory is enabled and the agent has a system message, the system message is preserved.

```typescript theme={null}
agent.clearConversationHistory();

const response = await agent.run({
  prompt: "Start a new conversation",
});
```

Use this after a user switches projects, changes accounts, or asks to start over.

## Next steps

<CardGroup cols={3}>
  <Card title="Structured Output" icon="brackets-curly" href="/typescript/agent/structured-output">
    Generate type-safe responses with Zod schemas.
  </Card>

  <Card title="Streaming" icon="wave-pulse" href="/typescript/agent/streaming">
    Stream agent responses in real time.
  </Card>
</CardGroup>