{"skill":{"slug":"openmemo-clawhub-skill","displayName":"OpenMemo Memory – Persistent Memory for OpenClaw Agents","summary":"Provides OpenClaw agents with local, scene-aware, persistent structured memory for task deduplication and long-term workflow recall.","description":"# OpenMemo - Persistent Memory for OpenClaw Agents\n\nStop agents from repeating tasks. Give your AI long-term memory.\n\n## The Problem\n\nOpenClaw provides a basic memory system, but in real-world usage agents still:\n\n- **Repeat the same tasks** — the agent deploys successfully, but runs the entire workflow again next time because it never recorded the result\n- **Store memory as large documents** — chat history and MEMORY.md files help retrieve text, but agents also need to remember tasks they completed, decisions they made, and workflows that succeeded\n\n## What OpenMemo Adds\n\nOpenMemo introduces a **structured memory layer** designed for agent workflows. Instead of storing raw conversation text, OpenMemo records **experience events**.\n\n```\nBackend deployed using Docker Compose\nScene: deployment\nType: task_execution\n```\n\nAgents recall **actions and results**, not just text.\n\n## Comparison\n\n| Feature | Typical Long-Term Memory | OpenMemo Memory |\n|---|---|---|\n| Memory type | Document memory | Experience memory |\n| Storage | Notes and logs | Structured events |\n| Retrieval | Vector search | Scene + task recall |\n| Task deduplication | No | Yes |\n| Workflow reuse | No | Yes |\n\n## Core Capabilities\n\n### Persistent Memory\n\nOpenMemo records structured experience from agent workflows: tasks completed, decisions made, workflows validated. These memories persist across sessions and can be recalled when similar tasks appear. Over time the agent accumulates **long-term operational knowledge**.\n\n### Task Deduplication\n\nOpenMemo introduces **task fingerprinting**. Before executing a task, the agent checks memory. If the task already succeeded, the agent can reuse the result, skip execution, or continue from the previous step. This prevents duplicate execution, wasted tokens, and repeated workflows.\n\n### Scene-Aware Memory\n\nOpenMemo detects the working context: `coding`, `research`, `debugging`, `deployment`. Only the most relevant memories are retrieved for the current task. This keeps context focused and efficient.\n\n### Memory Inspector\n\nA built-in dashboard lets you see what the agent remembers, memory ranking and recall results, and system health. The memory system becomes **transparent** instead of a black box.\n\n### Local-First Architecture\n\nAll memory operations happen locally. No external dependencies, no cloud required, full privacy, lower latency.\n\n```\nOpenClaw Agent\n      |\n      v\nOpenMemo Skill\n      |\n      v\nOpenMemo Adapter (local)\n      |\n      v\nOpenMemo Memory Engine\n```\n\n## Example\n\n**Without OpenMemo:**\n\n```\n> deploy backend\n  → agent rebuilds everything again\n```\n\n**With OpenMemo:**\n\n```\n> deploy backend\n  → agent detects previous deployment\n  → reuses workflow\n```\n\nThe agent stops behaving like a script and starts behaving like a **system**.\n\n## Tools\n\n### recall_memory\n\nRetrieve relevant memory from OpenMemo. Use this to recall past experience, decisions, and knowledge before executing tasks.\n\n**Parameters:**\n- `query` (string, required): The search query for memory recall\n- `scene` (string, optional): Scene context (e.g., coding, debug, research, deployment)\n\n### write_memory\n\nStore structured memory event in OpenMemo. Use this after completing important tasks to save experience for future use.\n\n**Parameters:**\n- `content` (string, required): The memory content to store\n- `scene` (string, optional): Scene context\n- `type` (string, optional): Memory type — fact, decision, observation, preference\n\n### check_task_memory\n\nCheck if a task has already been executed. Use this FIRST before starting any task to avoid duplication.\n\n**Parameters:**\n- `task_description` (string, required): Description of the task to check\n\n## Rules\n\nWhen executing tasks, follow these memory operating rules:\n\n1. BEFORE starting any task, call `check_task_memory` with the task description. If a successful previous execution exists, reuse the result or skip.\n\n2. Use `recall_memory` to retrieve relevant past experience before making decisions.\n\n3. After completing important tasks, call `write_memory` to store structured experience: decisions made, successful approaches, errors resolved, key observations.\n\n4. Always include the scene context (coding, debug, research, deployment) for better recall accuracy.\n\n## Setup\n\nInstall the OpenMemo adapter locally:\n\n```\npip install openmemo openmemo-openclaw\nopenmemo serve\n```\n\nRestart your agent. The Skill will automatically detect the adapter and activate persistent memory.\n\n## Best Use Cases\n\n- Coding agents\n- DevOps automation\n- Research agents\n- Multi-step AI workflows\n\n## Links\n\n- GitHub: https://github.com/openmemoai/openmemo\n- Adapter: https://github.com/openmemoai/openmemo-openclaw-adapter\n","tags":{"latest":"0.1.0"},"stats":{"comments":0,"downloads":719,"installsAllTime":0,"installsCurrent":0,"stars":1,"versions":1},"createdAt":1773213545718,"updatedAt":1778491824366},"latestVersion":{"version":"0.1.0","createdAt":1773213545718,"changelog":"OpenMemo 0.1.0 introduces structured, persistent memory for OpenClaw agents.\n\n- Adds a new memory layer that records structured \"experience events\" instead of only raw text.\n- Enables agents to remember and deduplicate tasks, preventing redundant executions and workflow repetition.\n- Introduces scene-aware memory to ensure agents recall only the most relevant context (coding, research, debugging, deployment).\n- Provides tools for memory recall, writing new memories, and checking if tasks have been executed.\n- Offers a transparent memory inspector dashboard and supports full local operation with no external dependencies.","license":"MIT-0"},"metadata":null,"owner":{"handle":"openmemoai","userId":"s1777zhh30mr5peynapyhaxnms83hzvg","displayName":"openmemoai","image":"https://avatars.githubusercontent.com/u/46662906?v=4"},"moderation":null}