Memory OS

OpenClaw Integration

Memory OS provides persistent memory for AI agents on the OpenClaw and ClawHub platform. Your agent can store memories, search semantically, and maintain context across sessions - all via a simple API or MCP tools.

What Your Agent Gets

CapabilityDescription
Persistent storageMemories survive across sessions and conversations
Semantic searchFind relevant memories by meaning, not just keywords
LLM contextPre-formatted context ready to inject into prompts
Memory tiersShort, medium, and long-term retention control
Metadata & tagsOrganize memories with custom metadata
Entity extractionTrack people, places, and concepts across memories

Agent Self-Signup

Your agent can create its own account and get an API key in a single request:

Bash
curl -X POST https://api.mymemoryos.com/api/v1/agent-signup \
  -H "Content-Type: application/json" \
  -d '{
    "agent_name": "my-openclaw-agent",
    "agent_description": "A research assistant that remembers context",
    "contact_email": "owner@example.com",
    "platform": "openclaw"
  }'

Response:

JSON
{
  "data": {
    "tenant_id": "uuid-here",
    "api_key": "mos_live_xxxxx",
    "tools_available": [
      "store_memory",
      "search_memories",
      "get_context",
      "get_memory",
      "list_memories",
      "delete_memory"
    ],
    "mcp_config": {
      "command": "npx",
      "args": ["-y", "@thriveventurelabs/memoryos-mcp"],
      "env": {
        "MEMORYOS_API_KEY": "mos_live_xxxxx"
      }
    }
  }
}
FieldRequiredDescription
agent_nameYesName of your agent
agent_descriptionNoWhat your agent does
contact_emailNoHuman owner can claim the account later
platformNoDefaults to "openclaw"

MCP Tools

Memory OS exposes 6 tools via the MCP server. Install with:

Bash
clawhub install memoryos

Or configure directly:

JSON
{
  "mcpServers": {
    "memoryos": {
      "command": "npx",
      "args": ["-y", "@thriveventurelabs/memoryos-mcp"],
      "env": {
        "MEMORYOS_API_KEY": "mos_live_xxxxx"
      }
    }
  }
}

Available Tools

ToolDescription
store_memoryStore a new memory with content, tier, and metadata
search_memoriesSemantic search across stored memories
get_contextGet relevant context formatted for LLM prompts
get_memoryRetrieve a specific memory by ID
list_memoriesList stored memories with optional filters
delete_memoryDelete a memory by ID

Agent Workflow Examples

Daily Journaling Agent

An agent that logs daily observations and recalls them when relevant:

JavaScript
// At the end of each day, store a summary
await memoryos.store_memory({
  content: "User completed 3 code reviews today. Mentioned being tired of reviewing auth PRs.",
  tier: "medium",
  importance: 0.6,
  metadata: { type: "daily_summary", date: "2025-01-15" }
});

// Next day, retrieve context before responding
const context = await memoryos.get_context({
  query: "What has the user been working on?",
  max_tokens: 2000
});

Cross-Session Context Agent

An agent that maintains continuity across conversations:

JavaScript
// Store key facts from the current session
await memoryos.store_memory({
  content: "User is building a React dashboard for their e-commerce platform. Using Tailwind CSS and Supabase.",
  tier: "long",
  importance: 0.8
});

// In a new session, search for project context
const results = await memoryos.search_memories({
  query: "React dashboard project",
  limit: 5,
  threshold: 0.5
});

Preference Learning Agent

An agent that remembers and applies user preferences:

JavaScript
// When user expresses a preference
await memoryos.store_memory({
  content: "User prefers TypeScript over JavaScript. Always provide TS examples first.",
  tier: "long",
  importance: 0.9,
  metadata: { type: "preference", category: "coding" }
});

// Before generating code examples
const prefs = await memoryos.search_memories({
  query: "coding preferences language",
  tier: "long",
  limit: 10
});

REST API Alternative

If your agent does not support MCP, use the REST API directly:

Bash
# Store a memory
curl -X POST https://api.mymemoryos.com/api/v1/memories \
  -H "Authorization: Bearer mos_live_xxxxx" \
  -H "Content-Type: application/json" \
  -d '{
    "content": "User prefers dark mode",
    "tier": "long",
    "importance_score": 0.7
  }'

# Search memories
curl -X POST https://api.mymemoryos.com/api/v1/search \
  -H "Authorization: Bearer mos_live_xxxxx" \
  -H "Content-Type: application/json" \
  -d '{
    "query": "user preferences",
    "limit": 10
  }'

# Get context for LLM
curl -X POST https://api.mymemoryos.com/api/v1/context \
  -H "Authorization: Bearer mos_live_xxxxx" \
  -H "Content-Type: application/json" \
  -d '{
    "query": "What does the user like?",
    "max_tokens": 2000
  }'

Scopes

Agent API keys are created with the following scopes:

ScopeAccess
memories:readRead and list memories
memories:writeCreate, update, and delete memories
search:readSemantic search and context retrieval

These cover all 6 MCP tools. Admin operations (key management, tenant settings) require upgrading via the dashboard.

Agent Memory Patterns

Advanced memory patterns for AI agents

MCP Server

Full MCP server documentation

API Authentication

API key scopes and authentication

Search API

Semantic search reference

Ctrl+Shift+C to copy