← 返回
未分类 中文

Plugin

End-to-end encrypted, decentralized memory for OpenClaw. Set up an account once, then call totalreclaw_remember / totalreclaw_recall (or the tr CLI under hyb...
端到端加密的去中心化记忆(OpenClaw),一次设置账户后调用 totalreclaw_remember / totalreclaw_recall(或 hyb 环境下的 tr CLI)即可。
p-diogo p-diogo 来源
未分类 clawhub v3.3.12-rc.12 8 版本 99781.8 Key: 无需
★ 0
Stars
📥 2,286
下载
💾 0
安装
8
版本
#agent-memory#e2e-encryption#e2ee#encryption#latest#memory#persistent-context#privacy#rc

概述

TotalReclaw — OpenClaw plugin

End-to-end encrypted memory stored across a decentralized network (not on a single company server). The plugin handles install + account-setup autonomously; this skill tells the agent what to call and when.

TOP RULE — MEMORY STORAGE

When the user states ANYTHING about themselves — preferences, decisions, identity, work, location, plans, tools, commitments — call totalreclaw_remember (or tr remember --json "" under hybrid-primary). NEVER write user facts to MEMORY.md, USER.md, or any local file. Those are agent-persona configuration; user memories live encrypted on-chain.

Trigger immediately, no "should I?": "I prefer / like / love / hate X", "my favorite is X", "I'm X", "my name is X", "I live / work at X", "I use X", "I decided X", "I'm going to X", "from now on X", "we agreed X", "I'll X by Y", "remind me to X", "remember that X", "save X", "note X". If a message contains MULTIPLE facts, call totalreclaw_remember once per fact — never blob.

When the user asks about themselves ("what's my X?", "do I X?", "what do you remember?"), call totalreclaw_recall FIRST. If 0 results, say so plainly — don't guess from session history.

Do not describe storage as "local", "local-only", "stored locally", "on disk", "local memory", "local storage", or "without an account" — memories are decentralized on-chain (not on any single company server) and an account is required to derive the encryption key.

Phrase safety (HARD)

The 12-word recovery phrase MUST NEVER enter chat or the agent's LLM context. The plugin's browser-based pair flow handles phrase generation and import. Never echo it, never ask the user to paste it, never pass it to any tool. If the user pastes one anyway, tell them it's compromised and re-pair for a fresh wallet.

Install

openclaw plugins install @totalreclaw/totalreclaw@rc
openclaw skills install totalreclaw

Plugin is published on npm + ClawHub (p-diogo, GitHub-verified). Stable: drop @rc. Under hybrid-primary, the tr CLI is the primary path; native totalreclaw_* tools are also bound when OpenClaw exposes them — use whichever your runtime binds.

Setup (plugin-driven)

The plugin auto-creates a pair session on first load when no credentials exist, writes {url, pin, sid, expires_at_ms} to ~/.totalreclaw/.pair-pending.json, and injects a before_agent_start context block to you. When you see that block, surface the URL and PIN to the user VERBATIM — read from the injected context; never invent values. Confirm on pair completion: ✓ TotalReclaw set up. (If credentials already exist: ✓ TotalReclaw is already set up.)

If a restart is needed (rare — plugin self-handles via gateway config-watcher), issue the literal slash command /totalreclaw-restart. Never kill, docker restart, or prompt the user to restart.

Hybrid-primary CLI

OpenClaw 2026.5.x may strip non-bundled tool bindings (issue #223). Never rely on totalreclaw_* tool binding alone. Fall back via shell:

TR_CLI="$(ls -d ~/.openclaw/npm/node_modules/@totalreclaw/totalreclaw/dist/tr-cli.js ~/.openclaw/extensions/totalreclaw/dist/tr-cli.js 2>/dev/null | head -1)"
node "$TR_CLI" status --json     # {"version":"...","hybrid_mode":true,"tool_count":17}
node "$TR_CLI" remember --json "I work at Graph Foundation"
node "$TR_CLI" recall --json "where does the user work" --limit 5   # {"results":[{"text":"...","score":0.8}]}
node "$TR_CLI" pair --json       # {"url":"...","pin":"123456","expires_at_ms":...} — plugin's auto-pair is preferred; only use this for re-pair

Pair is UNCONDITIONAL when no credentials — pasting the install URL IS the consent. tr --help lists every subcommand.

Tools (16 native; same surface via tr)

totalreclaw_remember · totalreclaw_recall · totalreclaw_forget · totalreclaw_pair · totalreclaw_pin · totalreclaw_unpin · totalreclaw_retype · totalreclaw_set_scope · totalreclaw_export · totalreclaw_import_from · totalreclaw_import_batch · totalreclaw_consolidate · totalreclaw_status · totalreclaw_upgrade · totalreclaw_onboarding_start · totalreclaw_report_qa_bug (RC only).

Full guides:

版本历史

共 8 个版本

  • v3.3.12-rc.12 当前
    2026-05-31 12:44
  • v3.3.12-rc.11
    2026-05-21 12:13 安全 安全
  • v3.3.12-rc.9
    2026-05-12 04:25 安全 安全
  • v3.3.12-rc.7
    2026-05-11 03:57 安全
  • v3.3.12-rc.4
    2026-05-09 15:57 安全
  • v3.3.11-rc.2
    2026-05-07 03:17 安全
  • v3.3.6-rc.1
    2026-05-03 02:53 安全
  • v3.3.2
    2026-04-30 15:57 安全

安全检测

腾讯云安全 (Keen)

队列中

腾讯云安全 (Sanbu)

队列中

🔗 相关推荐

ai-agent

Agent Browser

rez0
用于 AI 代理的浏览器自动化 CLI。当用户需要与网站交互(包括浏览页面、填写表单、点击按钮、截图等)时使用。
★ 834 📥 305,420
ai-agent

self-improving agent

pskoett
捕获经验教训、错误及修正内容,以实现持续改进。适用于以下场景:(1)命令或操作意外失败;(2)用户纠正Claude(如“不,那不对……”“实际上……”);(3)用户请求的功能不存在;(4)外部API或工具出现故障;(5)Claude发现自身
★ 4,098 📥 825,084
ai-agent

Self-Improving + Proactive Agent

ivangdavila
自我反思+自我批评+自我学习+自组织记忆。智能体评估自身工作、发现错误并持续改进。
★ 1,394 📥 322,372