OpenClaw Plugin

Sync local markdown memories into Echo cloud and retrieve them through OpenClaw

The OpenClaw plugin is the markdown-first integration. It scans local memory files, syncs them to Echo cloud, exposes retrieval commands, and provides a localhost UI for browsing the local archive.

What the plugin does

  • Scans markdown files from the configured OpenClaw memory directory.
  • Syncs content into Echo cloud.
  • Exposes manual commands for status, sync, search, graph access, and onboarding.
  • Registers memory retrieval tooling so OpenClaw can use EchoMemory during normal chat.

Plugin vs skill

OpenClaw plugin

The plugin is the runtime integration. It loads inside OpenClaw, syncs markdown memories, exposes /echo-memory commands, starts the local UI, and registers retrieval tools for normal chat.

OpenClaw skill

The skill is a guidance layer. It helps users install, configure, and troubleshoot the plugin, but it does not replace the plugin runtime and does not perform sync or retrieval by itself.

If you want the actual EchoMemory integration, install the plugin. If you want setup help inside OpenClaw, add the skill as a companion layer.

Before you begin

  1. Create or sign in to your EchoMemory account.
  2. Generate an ec_... API key.
  3. Set tools.profile to full in ~/.openclaw/openclaw.json.
  4. Know the absolute path to the markdown memory directory you want OpenClaw to sync.

Install the plugin

Public installs should come from the published plugin sources, not from a local GitHub clone.

cd $HOME\.openclaw npm install @echomem/openclaw-memory
  1. Open the published plugin listing in the OpenClaw Marketplace if you want the official listing and metadata.
  2. Install the published package inside ~/.openclaw using the exact scoped NPM package shown above.
  3. Restart openclaw gateway.
  4. Keep the plugin entry in openclaw.json.
Do not rely on a bare plugin name for EchoMemory on newer OpenClaw hosts. Use the exact scoped package or the published marketplace listing. The GitHub repo is mainly for source inspection or active development.

Configure the plugin

The only strictly required value is apiKey. In practice, most setups also provide memoryDir so the sync target is explicit.

{ "tools": { "profile": "full" }, "plugins": { "entries": { "@echomem/openclaw-memory": { "enabled": true, "config": { "apiKey": "ec_your_key_here", "memoryDir": "C:\Users\your-user\.openclaw\workspace\memory", "autoSync": false, "syncIntervalMinutes": 15 } } } } }

Once the plugin is configured and the gateway is restarted, these are the main commands users should expect to use during setup validation and day-to-day retrieval.

  • /echo-memory whoami
  • /echo-memory status
  • /echo-memory sync
  • /echo-memory search <query>
  • /echo-memory view

Verify and browse the local UI

A successful startup should show the localhost viewer URL and normal sync startup logs.

The local UI exposes the same setup flow in the left sidebar. The quick setup panel explains the onboarding sequence, while the configuration panel lets users save the active API key, memory directory, autosync, and retrieval behavior without editing raw JSON first.

Local UI overviewThe main workspace combines local markdown clustering with search, date filters, view mode, setup access, cloud tabs, sync selection, and direct graph launch in one surface.
Quick setup sidebarGives users the signup, API key, and OpenClaw profile steps in the same place they manage the archive.
Configuration sidebarLets users save cloud credentials, memory directory, autosync, timeout, and Echo-only retrieval settings locally.
Cloud memory tabShows synced memories, search inside the cloud index, and lets users compare cloud recall against the local markdown canvas.
Cloud source tabLists uploaded sources separately from memories so users can inspect what has actually been synced to Echo cloud.

Local UI

The localhost UI is for browsing local markdown files directly. It can stay useful even if you are not fully connected to Echo cloud yet.

Memory graph links

Graph commands open the yeahecho.com memory graph surfaces. They are not the same as the localhost local UI and should be documented separately.

Search and chat behavior

  • Manual search is deterministic through the explicit command path.
  • Normal chat retrieval is model-driven and uses the registered memory tool when the model decides it is relevant.
  • Semantic retrieval usually works better than literal exact-phrase matching.
  • If prompt injection is disabled in OpenClaw plugin settings, manual search still works but automatic retrieval is weaker.