AI Agent Memory System

Teach your AI Agent to remember. Learn how memories work, how to manage them, and how to use them for personalized workflow execution.

AI Agent Memory System

The FlowDot AI Agent can remember information across conversations. This makes interactions more natural and enables personalized workflow execution.

What You'll Learn

Part 1: Understanding Agent Memory

How It Works

  1. You tell the Agent something
  2. A memory model analyzes your message
  3. Important information is extracted
  4. Memories are stored in your profile
  5. Future conversations access these memories

What Gets Remembered

The Agent learns:

Part 2: Adding Memories

Explicit Memory Requests

Tell the Agent to remember:

You: Remember that I am the Demo User and I make YouTube videos.

The Agent:

  1. Acknowledges your request
  2. Extracts key information
  3. Shows a memory indicator (top right)
  4. You can approve or modify

Implicit Learning

The Agent also learns from context:

You: Check the weather in Buffalo, NY. That's where I live.

The Agent may remember:

Memory Indicator

When memories are created:

Part 3: Managing Memories

Viewing Your Memories

  1. Click the hamburger menu
  2. Go to Profile
  3. Find the Memories section

Memory Management

Action How
View all Profile > Memories
Edit Click the edit icon next to a memory
Delete Click the delete icon
Add manually Click "Add Memory" button

Asking the Agent to Forget

You: Forget everything you know about my location.

The Agent:

  1. Identifies relevant memories
  2. Asks for confirmation
  3. Deletes the specified memories

Part 4: Memory-Powered Workflow Execution

The Power of Context

Without memory:

You: What's the weather?
Agent: I need to know your location. Where are you?

With memory:

You: What's the weather?
Agent: [Remembers you live in Buffalo]
Agent: [Runs weather workflow for Buffalo]
Agent: It's 45°F and cloudy in Buffalo.

Example: Smart Weather

  1. Setup memory:

    You: Remember that I live near Buffalo, NY.

  2. Later conversation:

    You: Tell me the weather.
Agent: [Uses Buffalo as location]
[Runs weather workflow]
Returns local weather!

Example: Discord Integration

  1. Setup memory:

    You: Remember that my Discord channel ID is #123456.

  2. Later:

    You: Post the weather report to my Discord.
Agent: [Uses remembered channel ID]
[Runs Discord post workflow]
Done!

Part 5: Memory Configuration

Memory Model Selection

Configure which AI model manages memories:

  1. Go to hamburger menu > API Settings
  2. Find Preferred Models tab
  3. Look for Memory Manager setting
  4. Select your preferred model

Recommendation: Use a capable model for memory management. It needs to understand context and extract relevant information.

Enabling/Disabling Memory

Toggle the entire memory system:

  1. Go to the Agent interface
  2. Find the Memory toggle (usually top area)
  3. Toggle Off to disable
  4. Toggle On to re-enable

Your memories persist even when disabled.

Part 6: Memory Best Practices

Good Memory Subjects

Store Examples
Location City, timezone, country
Preferences Units (Celsius/Fahrenheit), language
Identifiers Discord channel, Slack workspace
Context Job role, interests, common tasks
Defaults Preferred models, output formats

Avoid Storing

Don't Store Why
Passwords Security risk
API keys Use User API Keys instead
Sensitive data Memory isn't encrypted
Temporary info Clutters memory space

Memory Hygiene

Periodically review and clean:

  1. Visit Profile > Memories
  2. Delete outdated information
  3. Correct inaccurate memories
  4. Consolidate similar memories

Part 7: Advanced Memory Use Cases

Business Context

You: Remember that I run a coffee shop called "Morning Brew" in Seattle.

Now the Agent can:

Workflow Preferences

You: Remember that when I ask for weather, I want a summary, not raw data.

The Agent learns your output preferences.

Multi-Workflow Context

You: Remember that my standard flow is: weather check, then calendar review, then news summary.

Now:

You: Do my morning routine.
Agent: [Executes all three workflows in order]

Part 8: Troubleshooting

"Agent doesn't remember"

  1. Check memory is enabled (toggle)
  2. Verify memory was saved (check Profile > Memories)
  3. Memory may have been auto-denied
  4. Try explicitly: "Remember that..."

"Agent remembers wrong info"

  1. Go to Profile > Memories
  2. Find and edit the incorrect memory
  3. Or ask: "Forget [incorrect info]"

"Too many memory prompts"

The Agent may over-extract memories:

  1. Deny unnecessary memory suggestions
  2. Use a more capable memory model
  3. Be explicit: "Don't remember this, just answer"

Summary

Agent memory creates personalized experiences:

Memories + Workflows = Truly personalized AI automation.

Next, explore building custom React frontends with FlowDot Apps!

Related Tutorials

Back to Tutorials