🐾Persistent memory for your personal AI

Your OpenClaw agent starts fresh every single morning

OpenClaw runs on your devices, answering your messages and managing your life — but without persistent memory, it forgets who you are between sessions. Your preferences, your people, your routines. Gone. AgentRecall gives your agent a memory that actually lasts.

Give Your Agent Memory →Compare Solutions

Persistent

User Preferences

Graph

Relationships

Cross-Channel

Memory

Learning

Daily Routines

Memory that actually knows you

Built for personal AI assistants that run on your devices. Not a generic key-value store — a memory system designed for how people actually use their agents.

💜

User Preference Memory

Your agent remembers how you like things — dark mode or light, your dietary restrictions, your communication style, the way you like your calendar organized. No more re-explaining yourself every session.

// Your agent stores your preferences
await recall.store({
  content: "User prefers dark mode, concise replies,
  and gluten-free meals when ordering food",
  agentId: "openclaw",
  category: "preference"
});

// Next time it just knows.

🕸️

Relationship Graph

Your agent builds a graph of everyone in your life — who they are, how they connect, what they care about. Ask "tell me about Sarah" and it knows she's your sister, her birthday is March 12, and she's allergic to peanuts.

// Your agent learns relationships
await recall.store({
  content: "Sarah is my sister. Her birthday is
  March 12. She's allergic to peanuts
  and works at Acme Corp.",
  agentId: "openclaw"
});

// Entities auto-extracted:
// → Sarah → sibling_of → User
// → Sarah → works_at → Acme Corp
// → Sarah → allergy → peanuts

🔄

Daily Routine Learning

AgentRecall observes your patterns and learns your routines. It knows you check emails at 9am, have standup on Tuesdays, and prefer to batch errands on Saturday mornings. Predictive, not reactive.

// Agent learns your daily patterns
await recall.store({
  content: "User has standup every Tuesday and
  Thursday at 10am. Prefers not to
  schedule meetings before 11am
  on Mondays.",
  agentId: "openclaw",
  category: "temporal"
});

// Smart scheduling suggestions

📱

Cross-Channel Memory

You talk to your agent on WhatsApp, Telegram, and Discord. With AgentRecall, the context follows you everywhere. That decision you made on Telegram? Your agent knows about it in Discord too.

// Same memory, all channels
await recall.search({
  query: "what did we discuss about the
  vacation plans?",
  agentId: "openclaw",
  channel: "cross"
});

// Finds conversations from WhatsApp,
// Telegram, Discord — all unified

One command. Your agent remembers everything.

Connect AgentRecall to OpenClaw via MCP. No code changes needed — your agent gains persistent memory instantly.

# Install the MCP server
pip install agentrecall-mcp

# Connect to your OpenClaw agent
openclaw mcp set agentrecall '{"command":"agentrecall-mcp","env":{"AGENTRECALL_API_KEY":"ark_YOUR_KEY"}}'

# That's it. Restart your agent and it remembers.
# Your preferences, relationships, and routines
# persist across every session.

How does it compare?

AgentRecall isn't the only option. Here's why it's the right one for your personal AI.

Feature	AgentRecall	Alternatives
Remembers user preferences across sessions	✓	✗
Graph relationships between people & things	✓	✗
Cross-channel memory (WhatsApp + Telegram + Discord)	✓	✗
AI-powered memory processing (auto-categorize)	✓	✗
Semantic search by meaning, not keywords	✓	✓
Runs locally on your device (BYOK)	✓	✓
Learns daily routines over time	✓	✗
Open source & free tier	✓	✓