Openclawdy

OpenClawdy

Memory Infrastructure for Autonomous Agents

Give your agent persistent memory that survives sessions. Store facts, preferences, decisions, and learnings - recall them semantically whenever needed. Advanced features include reputation tracking, cross-agent memory pools, and time-travel snapshots.

API Host & Protocol

Installation

openclaw skill install openclawdy

Or add to your agent config:

skills:
  - url: https://openclawdy.xyz/SKILL.md
    name: openclawdy

Authentication & Security

OpenClawdy uses wallet-based authentication with message signing only.

How It Works

1. Your agent signs a timestamp message with its wallet
2. The signature + address are sent in request headers
3. Server verifies the signature (no private key ever leaves your agent)

Required Headers

X-Agent-Address: 0x...      # Your wallet address (public)
X-Agent-Signature: 0x...    # Signed message (proves ownership)
X-Agent-Timestamp: 123...   # Unix timestamp (ms, prevents replay)

Message Format

OpenClawdy Auth
Timestamp: {timestamp}

Security Guarantees

• No private key access required - Only signing capability needed
• Wallet isolation - Each wallet gets its own isolated memory vault
• No env vars needed - Authentication is header-based
• No stored credentials - Signatures are verified per-request

Privacy & Data Policy

Core Tools

memory_store

Store information for later retrieval.

Endpoint: POST /api/memory/store

Parameters:

• content (required): The information to remember
• type (optional): Category of memory - one of: fact, preference, decision, learning, history, context. Default: fact
• tags (optional): Array of tags for organization

Example Request:

{
  "content": "User prefers TypeScript over JavaScript",
  "type": "preference",
  "tags": ["coding", "language"]
}

Example Response:

{
  "success": true,
  "data": {
    "id": "mem_abc123",
    "content": "User prefers TypeScript over JavaScript",
    "type": "preference",
    "tags": ["coding", "language"],
    "createdAt": "2025-02-10T12:00:00Z"
  }
}

memory_recall

Retrieve relevant memories using semantic search. Finds memories by meaning, not just keywords.

Endpoint: POST /api/memory/recall

Parameters:

• query (required): What to search for
• limit (optional): Maximum results to return (1-20). Default: 5
• type (optional): Filter by memory type

Example Request:

{
  "query": "programming language preferences",
  "limit": 3
}

Example Response:

{
  "success": true,
  "data": [
    {
      "id": "mem_abc123",
      "content": "User prefers TypeScript over JavaScript",
      "type": "preference",
      "relevance": 0.95,
      "createdAt": "2025-02-10T12:00:00Z"
    }
  ]
}

memory_list

List recent memories without semantic search.

Endpoint: GET /api/memory/list

Parameters:

• type (optional): Filter by memory type
• limit (optional): Maximum results (1-100). Default: 20
• offset (optional): Pagination offset. Default: 0

memory_delete

Delete a specific memory by ID.

Endpoint: DELETE /api/memory/{id}

Parameters:

• id (required): The memory ID to delete

memory_clear

Clear all memories in the vault. Use with caution - this is irreversible.

Endpoint: DELETE /api/memory/vault

memory_export

Export all memories as JSON for backup.

Endpoint: GET /api/memory/vault

memory_stats

Get usage statistics for your agent.

Endpoint: GET /api/agent/stats

Example Response:

{
  "success": true,
  "data": {
    "address": "0x1234...",
    "tier": "free",
    "memoriesStored": 150,
    "recallsToday": 45,
    "limits": {
      "maxMemories": 1000,
      "maxRecallsPerDay": 100
    }
  }
}

Advanced Tools

memory_reputation

Track which memories lead to good outcomes. Store memories with reputation scores, update based on success/failure, recall memories ranked by proven effectiveness.

Endpoints:

• POST /api/memory/reputation/store - Store with reputation
• POST /api/memory/reputation/recall - Recall by reputation rank
• POST /api/memory/reputation/update - Update reputation score

store_ranked

Request:

{
  "content": "Use retry logic for API calls",
  "type": "learning",
  "reputation": 0.8
}

recall_ranked

Request:

{
  "query": "error handling strategies"
}

Response:

{
  "success": true,
  "data": [
    {
      "id": "mem_xyz",
      "content": "Use exponential backoff for retries",
      "reputation": 0.92,
      "usage_count": 15,
      "success_rate": 0.93
    }
  ]
}

update_reputation

Request:

{
  "memory_id": "mem_xyz",
  "outcome": "success",
  "impact": 0.8
}

memory_pool

Cross-Agent Memory Pools - Share knowledge between multiple agents. Create pools, store shared memories, recall from collective intelligence. Perfect for agent teams and swarms.

Endpoints:

• POST /api/memory/pool/create - Create new pool
• POST /api/memory/pool/store - Store in pool
• POST /api/memory/pool/recall - Search pool
• GET /api/memory/pool/list - List accessible pools

create

Request:

{
  "name": "research-team"
}

Response:

{
  "success": true,
  "data": {
    "pool_id": "pool_abc123",
    "name": "research-team",
    "created_at": "2025-02-10T12:00:00Z"
  }
}

store

Request:

{
  "pool_id": "pool_abc123",
  "content": "Found bug in authentication module - fix applied",
  "type": "fact"
}

recall

Request:

{
  "pool_id": "pool_abc123",
  "query": "authentication issues"
}

memory_snapshot

Memory Time Travel - Snapshot and restore agent memory states. Debug decisions by viewing past states, compare memory changes, restore to previous checkpoints. Essential for high-stakes agents.

Endpoints:

• POST /api/memory/snapshot/create - Create snapshot
• POST /api/memory/snapshot/restore - Restore from snapshot
• GET /api/memory/snapshot/list - List snapshots
• POST /api/memory/snapshot/compare - Compare snapshots

create

Request:

{
  "name": "before-major-update"
}

Response:

{
  "success": true,
  "data": {
    "snapshot_id": "snap_abc123",
    "name": "before-major-update",
    "memory_count": 150,
    "created_at": "2025-02-10T12:00:00Z"
  }
}

restore

Request:

{
  "snapshot_id": "snap_abc123",
  "mode": "read_only"
}

Modes: read_only (view only) or overwrite (replace current state)

compare

Request:

{
  "snapshot_id": "snap_abc123",
  "compare_to": "current"
}

Response:

{
  "success": true,
  "data": {
    "added": 12,
    "removed": 3,
    "modified": 5,
    "unchanged": 130
  }
}

Memory Types

Rate Limits

Error Responses

All endpoints return consistent error format:

{
  "success": false,
  "error": "Error message here",
  "code": "ERROR_CODE"
}

ACP Integration

OpenClawdy is available on the Agent Commerce Protocol (ACP). Other agents can purchase memory services directly:

Support

• Website: https://openclawdy.xyz
• API Status: https://openclawdy.xyz/api/health
• Twitter: @openclawdy
• ACP Agent: OpenClawdy Memory

License

MIT