竞品对比与迁移指南 Myrm 被设计为完整的 AI Agent 工作空间。与仅面向终端的编程 Agent 不同，Myrm 提供 GUI 优先的体验、持久化沙箱、跨会话记忆与企业级安全——同时保留完整编程能力。
迁移信心 ——数据与节奏在你手里：可选 SaaS、自托管或桌面三模；Hermes 技能可 GUI 拖拽 ZIP 导入；先对照下方能力表与分竞品章节再决定。文档提供中英双语 ；官网会按你的语言自动打开对应版本。
能力 Myrm Hermes Agent (v0.15) OpenClaw 360 Security OpenClaw OpenClacky MiniMax Mavis MemPalace 记忆系统 ✅ 8 类 + 知识图谱 ⚠️ 2,200 字符 ⚠️ 3 类 ⚠️ 3 类 ❌ 仅会话 ⚠️ 基础更新 ⚠️ 扁平抽屉 对话搜索 ✅ FTS5 + Qdrant 混合 ⚠️ v0.15 仅本地文本 ❌ ❌ ❌ ❌ ❌ GUI 界面 ✅ Web 原生 ❌ CLI（v0.15 TUI 多会话） CLI ✅ Web/App ⚠️ 基础 WebUI ⚠️ 飞书嵌入 ❌ 仅 CLI 桌面应用 ✅ Tauri Electron ❌ ❌ ❌ ❌ ❌ 部署 ✅ Web/Tauri/SaaS 自托管 自托管 ❌ 仅 SaaS 本地 + WebUI ❌ 闭源 SaaS 仅本地 SaaS 登录 ✅ Google OAuth + 一次性交换（JWT 不进 URL）；企业 OIDC 就绪 ❌ ❌（OAuth 仅 LLM Key） ⚠️ 厂商账号 ❌ ❌ 闭源 SaaS ❌ 模型支持 ✅ 100+ 模型 20+ 固定 ⚠️ 固定三档 BYOK 多模型 ❌ 仅 MiniMax N/A 渠道 ✅ 25+ 消息（25 内置） ~23（19 带平台提示） ~7 ⚠️ 飞书/钉钉 ⚠️ IM 机器人 ❌ 仅飞书 ❌ 子 Agent ✅ 6 模式 + 动态工作流 ⚠️ v0.15 Kanban Swarm（1 模式） ⚠️ 基础 spawn（仅线性） ⚠️ 基础 spawn ❌ 单 Agent ⚠️ L-W-V ❌ 安全 ✅ 6 层防御 ⚠️ v0.15 Promptware 3 卡口 1 层 ⚠️ 未知 ⚠️ 基础沙箱 ⚠️ 闭源 ❌ 凭证库 ✅ 标签注入 + TOTP（浏览器 + 桌面） ❌ 工具参数明文 ❌ ❌ ❌ ❌ ❌ WebUI 访问安全 ✅ 零信任 WebSocket + 安全 Cookie ⚠️ 仅 HTTP，WS 脆弱 ⚠️ 基础 JWT ⚠️ 未知 ⚠️ 仅 HTTP ⚠️ 闭源 SaaS ❌ LAN 裸奔 目标模式 ✅ 7 状态 + 4D 预算 /goal ❌ ❌ ❌ ⚠️ 计划审批 ❌ 上下文管理 ✅ 6 层 Prompt Cache + 22+ 中间件流水线 ⚠️ system_and_3（4 断点）+ 8 层固定栈 LCM LCM ⚠️ 空闲压缩 + 双缓存 ⚠️ Worker 隔离 ⚠️ 4 层栈 智能并发路由 ✅ O(1) 路径锁，零读写竞态 ⚠️ 仅粗粒度写锁 ❌ 不安全 ⚠️ 未知 ❌ ⚠️ 未知 ❌ 无头无人值守 ✅ 标签工具隔离（零死锁） ⚠️ 易交互死锁 ❌ 不支持 ❌ ❌ ❌ ❌ 极端防爆 ✅ 4 层护城河（卫生/护盾/剥离/预算） ⚠️ 大负载崩溃/丢失 ⚠️ 单点失败 ⚠️ 未知 ⚠️ 基础限制 ⚠️ 未知 ❌ 错误恢复 ✅ 14 类自愈 try/catch 基础 基础 ⚠️ 上下文溢出恢复 ⚠️ 未知 ❌ 企业可靠性 ✅ xdist 锁、EventBus 截断、OTEL 安全、路径边界 + TOCTOU 审批测试；聊天 SSE 状态机回归（错误/审批/文件 diff） ⚠️ 仅快乐路径测试 ⚠️ 易竞态 ⚠️ 未知 ⚠️ 未知 ⚠️ 未知 ❌ 文件编辑安全 ✅ 6 层保护 ❌ ❌ ❌ ❌ ⚠️ 未知 ❌ 定时任务 ✅ GUI 配置 Cron ❌ ✅ ✅ ❌ ❌ ❌ 统一工具网关 ✅ 四合一网关 + 弹性 BYOK 回退 ⚠️ 硬切换（易出错） ❌ N/A ⚠️ 未知 ❌ ⚠️ 未知 ❌ 语音 ✅ STT + TTS + 实时 ❌ WebRTC ⚠️ 基础 ❌ ❌ ❌ Computer Use ✅ 桌面自动化 + BBox 审批 + Tauri OS 叠加层 ❌ ❌ ❌ ❌ ❌ ❌ 浏览器引擎 ✅ 三层引擎 ❌ ❌ ❌ ❌ ⚠️ 未知 ❌ 网页搜索 ✅ 7 引擎 + BM25/Reranker 过滤 ⚠️ 云 API 透传 ⚠️ DDG/Bing 抓取 ⚠️ 同 OpenClaw ⚠️ DDG/Bing ❌ ❌ 网页抓取 ✅ 三层 + 向量提取 + DOM 修剪 ⚠️ web_extract（Firecrawl + LLM） ⚠️ HTTP / Firecrawl 回退 ⚠️ 同上 ⚠️ 仅 HTTP ❌ ❌ 图片搜索 ✅ 内置 image_search ❌ ❌ ❌ ❌ ❌ ❌ 长文目录 ✅ 聊天内自动 TOC + 滚动同步 ❌ ❌ ❌ ❌ ❌ ❌ 精准多模态 ✅ 意图感知视觉（无 UI 操作 = 0 vision token） ⚠️ 常开（浪费 token） ⚠️ 盲（无截图） ⚠️ 未知 ⚠️ 盲 ⚠️ 盲 ❌ 标题生成 ✅ O(1) 防阻塞 + 脱敏 ⚠️ 易卡死/泄漏 ⚠️ 易卡死/泄漏 ⚠️ 未知 ⚠️ 未知 ⚠️ 未知 ❌ 全局截屏 AI ✅ 快捷键截屏 + OCR + FlowPad ❌ ❌ ❌ ❌ ❌ ❌ Token 效率 ✅ 总计 ~2,167 tokens（少 86%） ~15,520 ~18,000 ⚠️ 手动三档 ⚠️ 16 工具（LLM 压缩成本） ⚠️ 高成本 ⚠️ ~900 唤醒
vs 360 Security OpenClaw（企业封装版） 360 Security OpenClaw 是基于 OpenClaw 核心的闭源企业封装，通过「Shrimp Coach」引导配置和手动「Token 成本模式」（轻量/经济/全功率）降低上手门槛。
Myrm 的领先之处 领域 360 Security OpenClaw Myrm 用户收益 Agent 创建 「Shrimp Coach」对话 19+ 预设 Agent + GUI 向导 零提示词即开即用，点击即用。 Token 成本控制 手动三档切换 自动三档复杂度路由 系统自动选最便宜且够用的模型，无需猜档位。 成本可见性 基础统计 15+ GUI 仪表盘 实时单条消息成本、缓存节省、工具用量分解。 云环境 云电脑 沙箱 + 持久终端 + SaaS 真跨平台隔离与数据归属。 多媒体 基础视频 Agent 原生 video/generator.py + 全双工语音 创作内容并支持打断（barge-in）免提交互。
结论：Myrm 提供更自动化、原生的 GUI 体验。无需「教练」问答，直接给可用模板；无需手动猜成本档位，智能路由任务并提供约 10 倍成本透明度。 vs OpenClacky — Token 优化本地 Agent OpenClacky 主打低成本本地 AI Agent，宣称 Token 约为 Hermes 的 1/6。核心策略：16 个最小工具、空闲压缩、先插入后压缩、双缓存标记、BYOK 多模型路由。
Myrm 的领先之处 领域 OpenClacky Myrm 用户收益 Tool Architecture 16 core + invoke_skill (2-tier) 3-tier (CORE/COMMON/EXTENDED) + ASCS cognitive load scoring + Dynamic Schema Weaver Scientifically optimized, dynamic tool pruning prevents hallucinations Idle Pipeline Single-purpose message compression (Thread) 3 parallel tasks (memory consolidation + evidence mining + cache preheating) + MaintenanceScheduler + crash-resilient registryIdle time does 3x more work Compression LLM-based summarization (each compression costs LLM tokens) Rule engine 3-tier (Dedup/Truncate/Remove) — zero LLM cost + Cold Cache Drain + Anti-ThrashingNo extra LLM calls for compression Cache Strategy Fixed last-2-message marking Multi-breakpoint (system + 15-block protection + compression boundary + last message) + TTL strategy + 20-block window protection + token distance validationMore precise, provider-aware caching Model Routing Simple main/sub-task model split 4-layer routing (ComplexityRouter + PrivacyRouter + FallbackManager + KeyPool)Comprehensive cost + privacy + reliability routing Cache Protection None Hot Cache Bypass (skip compression when cache is warm) + Anti-Thrashing (stop after 2+ ineffective compressions)Unique — no competitor has this Deployment CLI + WebUI Web + Tauri + SaaS (3 modes)More deployment options Memory System Session-only (lost on restart) 8 memory types + knowledge graph + cross-session consolidation Persistent intelligence across sessions GUI Basic web terminal Full workspace GUI — agent templates, Kanban, memory panel, analyticsProfessional workspace vs terminal Security Basic sandboxing 6-layer defense with PII protection + taint tracking + auditEnterprise-grade security
关键架构差异 压缩成本 ：OpenClacky 调用 LLM 生成压缩摘要 — 每次压缩都产生 API 成本。Myrm 使用确定性规则引擎（Dedup → Truncate → Remove），零 LLM 开销。缓存智能 ：OpenClacky 双缓存标记是静态模式（始终标记最后 2 条消息）。Myrm 的 ExplicitCacheProcessor 根据内容块、token 距离与压缩边界动态计算断点，并带完整校验流水线。空闲利用 ：OpenClacky 空闲定时器仅压缩消息。Myrm 空闲流水线并行 3 项：认知记忆整合、会话证据挖掘（从失败中学习）、前缀缓存预热 — 由容量感知调度器协调，带熔断保护。从 OpenClacky 迁移到 Myrm OpenClacky 功能 Myrm 对应 体验 16 core tools 3-tier tool system + ASCS scoring ⬆️ Upgrade Idle compression Idle pipeline (3 tasks + scheduler) ⬆️ Upgrade Insert-then-Compress Rule engine compression (zero LLM cost) ⬆️ Upgrade Dual cache marking Multi-breakpoint strategy + validation ⬆️ Upgrade BYOK model routing 4-layer routing system ⬆️ Upgrade Session persistence 8-type memory + knowledge graph ⬆️ Upgrade WebUI Full workspace GUI (Tauri + Web + SaaS) ⬆️ Upgrade Skill system 42-module skill evolution with safety ⬆️ Upgrade
结论：8 项升级、0 项持平、0 项降级。OpenClacky 用户获得零成本压缩、智能缓存保护、跨会话持久记忆与完整 GUI 工作区，同时保留全部 Token 效率优势。 统一工具网关与弹性 BYOK（vs Hermes / OpenClaw） 竞品常迫使用户管理大量 API Key，或依赖僵硬、易出错的网关。Myrm 提供 四合一统一工具网关 与 弹性 BYOK 回退 机制。
Myrm 的领先之处
四合一网关 ：一次订阅解锁 LLM、网页搜索、图像生成与 TTS，开箱零配置。弹性 Try-Catch 回退 ：官方网关不可用或 Work Units (WU) 用尽时，Myrm 自动无缝降级到你本地配置的 API Key，业务不中断。配额结转 ：与传统 SaaS 未用配额作废不同，Myrm 允许未用 WU 结转至下月。可视化网关状态 ：专用 GUI 仪表盘展示各工具状态（网关托管 / 自定义 Key / 未配置），Pro 用户有智能提示启用网关能力。金融级安全 ：PAT 经安全 POST 体传输，严格正则校验，防 SSRF，保证 token 不进 URL 日志。极端网络韧性 ：命令式按需轮询 + AbortController 8 秒硬超时，消除竞品自动轮询仪表盘导致的 UI 卡死与 API 配额空耗。
竞品痛点
Hermes ：网关「非黑即白」硬切换；网关失败或额度用尽则 Agent 直接崩溃停工。OpenClaw ：完全依赖用户通过扩展提供的 Key，无统一网关与计费。云浏览器依赖 ：Hermes 网页任务依赖第三方云浏览器（如 Browser Use），高延迟与隐私风险。Myrm 坚持本地/自托管沙箱做浏览器自动化，零延迟、数据完全私有。
迁移收益
零 Key 管理 ：不再为 OpenAI、Firecrawl、FAL 等疲于切换 API Key。安心 ：弹性回退 = 托管网关便利 + 自有备份 Key 可靠性。
WebUI 安全与本地/远程分离 多数竞品的 Web 界面要么在公网桥接时裸奔暴露（LAN/隧道无保护），要么对单人本地开发过于沉重（localhost 也强制数据库与登录）。
Myrm 以奥卡姆剃刀式 WebUI 安全方案，解决本地优先部署的「最后一公里」。
Myrm 的领先之处
本地零摩擦 vs 远程铁壁 ：WebUI 自动识别 localhost 回环免登录；一经隧道或 LAN 暴露即强制密码保护。零信任 WebSocket 网关 ：竞品常只保护 HTTP，WebSocket（实时日志/语音）裸奔。Myrm WsAuthMiddleware 在 session cookie 缺失或无效时物理拒绝握手（HTTP 403）。即时会话驱逐 ：改密码或切换保护时自动轮换全局 HMAC Session Signing Key，瞬间作废所有会话（类似失窃设备一键熔断）；竞品旧 JWT 常仍有效。单租户 Vault vs 重型 DB ：本地模式用单一高加密 admin.json，非 SaaS 式 Postgres/MySQL + RBAC 膨胀；安全、可移植、零依赖。
多 Agent 编排：确定性调度（vs Hermes / OpenClaw） 竞品多依赖单一编排模式（如 Hermes Kanban Swarm 或 OpenClaw 基础线性 spawn）。Myrm 提供 6 模式确定性编排引擎 ，辅以 5 层工具安全围栏 与 4 维预算控制 。
Myrm 的领先之处 领域 Hermes / OpenClaw Myrm 用户收益 Orchestration Modes ⚠️ 1-2 modes (Kanban or Spawn) 6 Modes (DAG, Swarm, Chain, Batch, Verified, Spawn)Right tool for the right job; DAG guarantees execution order without relying on LLM improvisation. Tool Isolation ⚠️ Basic global scope 5-Layer Fence (Type, Blocklist, Config, Intersection, Role)Complete prevention of child agents escalating privileges or executing unauthorized tools. Budget Control ⚠️ Token limit only 4D Budget (Tokens, USD, Time, Descendants)Perfect predictability for cost and execution depth. Zero bill shock. Execution Verification ❌ None CompletionGuard (Evidence-based)Children must prove task completion with actual STDOUT/STDERR, not just say “I’m done”.
结论：Myrm 用确定性软件工程模式取代「提示并祈祷」式多 Agent。流水线每次可靠、安全、在预算内执行。 智能并发路由 — 消除读写竞态（vs Hermes / OpenClaw） 高并发 Agent 沙箱中，LLM 常幻觉出破坏数据的操作，例如在同一并行工具批次中「同时读文件 X 又写文件 X」。竞品架构会导致脏读与灾难性写覆盖。
Myrm 引入 智能并发路由 ，在中间件层主动拦截并重路由冲突操作。
Myrm 的领先之处 领域 Hermes Agent OpenClaw Myrm 用户收益 Race Condition Defense ⚠️ Write-only lock ❌ None Full read-write exclusion Prevents dirty reads and corrupted files Concurrency Degradation ⚠️ Basic serial ❌ Fails Auto-degrade to safe sequential Safe fallback without crashing Lock Granularity ⚠️ Coarse path ❌ None Deep directory & block-level fingerprinting Unrelated parallel tasks keep running at max speed Performance Overhead O(N) linear scan N/A O(1) path lock resolution Zero latency penalty for the engine
结论：Myrm 彻底消除 LLM 幻觉导致的文件级竞态。冲突批次优雅展开为顺序队列，无关并行任务零性能损失。 无头 Agent：零死锁后台任务（vs Hermes / OpenClaw） SaaS 定时任务（Cron）、批处理或后台自动化等无头环境中，Agent 无人监督运行。若 LLM 幻觉调用人工介入（HITL）工具（如提问或渲染 UI 表单），竞品会无限等待永不到来的用户输入，灾难性死锁并烧穿算力。
Myrm 在框架最底层提供 基于标签的环境降级 架构。
Myrm 的领先之处 领域 Hermes Agent OpenClaw Myrm 用户收益 Deadlock Prevention ❌ Prone to freeze ❌ Unsupported 100% Guaranteed Zero Deadlock Scheduled background tasks will never hang waiting for human input. Tool Stripping ⚠️ Hardcoded in business logic ❌ None Automatic via tags=["interactive"] The LLM doesn’t even know the tool exists, preventing hallucination errors. Ecosystem Compatibility ❌ Needs manual wrapper for each tool ❌ None Works with any MCP Tool natively Any third-party MCP tool tagged interactive is automatically isolated. Architectural Layer ⚠️ Server-level hack ❌ None Core Harness / Agent engine layer The protection travels with the engine, whether deployed locally, Desktop, or SaaS.
结论：Myrm 保证无人值守工作流防弹可靠。后台任务物理剥离交互工具，在 LLM 犯错前根除 Cron 死锁根因。 极端场景防爆（vs Hermes / OpenClaw） 自主多模态场景（如 Computer Use）或超长会话中，Agent 必然撞上下文窗口上限；竞品频繁 OOM、断连或记忆全丢。
Myrm 实现 四层极端防爆护城河 ，确保 Agent 不崩溃、对话上下文不丢。
Myrm 的领先之处 领域 Hermes / OpenClaw Myrm 用户收益 Gateway Hygiene Allows massive payloads to hit the LLM layer, leading to severe compute node OOMs. Millisecond Interception Blocks malformed/gigantic payloads (>120K tokens) at the API gateway before they even touch the agent engine. Auxiliary Model Shield Crashes the application if the summarizer model’s context window is smaller than the payload being compressed. Dynamic Ratio Validation & Graceful Degradation Automatically falls back to the main model for summarization if the cheap auxiliary model is too small. No crash, no data loss. Media Stripping Treats all images equally; long-tail screenshots from past turns quickly bloat the context window. Sliding Visual Evidence Window Retains only the last 2 turns of images for reasoning, stripping large Base64 encodings from all older history. Drastically saves tokens and prolongs context life. Tail Budget Protection Hardcoded turn truncation (e.g., keep last N messages) often chops off critical working memory. Token Budget Reservation Strict reservation (e.g., 20% of max context tokens) exclusively for the most recent tail, guaranteeing active tasks are never squeezed out.
结论：极端负载下仍丝滑。剥离历史图片大幅省 Token，不必担心 Agent 突然崩溃抹掉劳动成果。 网页搜索 + 网页抓取 — 双引擎（vs Hermes / OpenClaw / Claude Code） Myrm 将 web_search 与 web_fetch 作为一等内置工具。竞品要么缺失、要么透传原始 API 结果、要么按次向云 API 计费。
你将获得 能力 Technical 用户收益 Web Search 7 engines + 7 intent types (zero LLM) + BM25/RRF + optional Reranker Relevant snippets only — no duplicate junk in context Web Fetch HTTP → Browser → Stealth + DOM pruning + Markdown Article body, not nav/ads HTML — ~40–50% fewer tokens Logged-in fetch SessionVault: CookieJar (HTTP) + storage_state (browser) One login works across fetch tiers — competitors often lose cookies on engine switch fetch_and_extract Crawl → BM25 + vector hybrid → Reranker Long pages → only relevant passages (zero LLM cost vs Hermes Gemini summary) Image search Built-in image_search_tool Visual research without extra Skills Cost 100% local processing $0/month vs Firecrawl/Jina/Exa subscriptionsChinese search SearxNG + Baidu aggregation Works out of the box — no Tavily/baidu Skill hacks PTC inline tools.web_search / tools.web_fetch in sandboxComplex flows without extra API round-trips
竞品痛点 Product Search Fetch Problem Hermes Exa/Tavily/Firecrawl API passthrough web_extract + Gemini LLM summaryPaid APIs + extra LLM token per page OpenClaw / OpenClacky DuckDuckGo/Bing HTML scrape Single HTTP or Firecrawl fallback Poor Chinese results, rate limits, regex HTML strip OpenCode Exa/Parallel cloud only Single HTTP + Turndown (full page) Requires API keys; no local filter pipeline Claude Code Hosted web_search Hosted web_fetch No self-host, no engine choice, opaque filtering
:::note 诚实边界
Hermes web_extract 看似「零配置」因用 LLM 摘要而非本地 embedding — 但每页消耗 token。Myrm web_fetch 开箱本地 DOM 修剪；配置后可 fetch_and_extract 向量+Reranker。SSRF 防护相当 — Myrm 优势在 过滤流水线 ，非基础安全。
:::
迁移收益 From After migrating to Myrm Hermes Drop Firecrawl dependency; local 3-tier fetch; GUI search engine config OpenClaw No manual Tavily/baidu Skill; 7-engine fallback; BM25/Reranker filtering Claude Code Self-hosted + 100+ models; auditable filter pipeline; PTC-native web tools
配置与用法见网页搜索与抓取指南 。
vs Hermes Agent (v0.15 Velocity) — 多 Agent 平台 Hermes v0.15（Velocity，2026-05-28）将核心循环从 16k 行重构为 14 模块共 3.8k 行，新增 Kanban Swarm 编排（104 PR），session_search 重写为无 LLM（~20ms，原 ~30s），引入 Promptware 防御（3 卡口、~15 Brainworm/C2 模式）。显著「还技术债」版本。
Myrm 的领先之处
事件驱动看板 vs Hermes 轮询 — 即时分派 + 心跳 + 僵尸检测6 条诊断规则 + 严重度自动升级（warning→error→critical）— 检测滞留 Ready、重复失败、长期 Blocked、死依赖、Triage 停滞、Block↔Unblock 循环（每卡 O(1) vs Hermes O(N) 事件扫描）6 种编排模式 （Spawn/Chain/DAG/Batch/Verified/Swarm Fission）vs v0.15 单一 SwarmCompletionGuard 物理证据验证 — Hermes 无完成验证器流水线模板向导 （发现问题 + 角色自动匹配）— Hermes hermes kanban swarm 为固定 CLI35+ 消息渠道 （25 带输出提示）vs ~23（19 带平台提示）— Myrm 覆盖钉钉、Teams、Google Chat、LINE、IRC、iMessage、Voice、Webhook、Zalo 等 Hermes 缺失项8 类记忆 + 知识图谱 vs 2,200 字符扁平记忆（MEMORY.md + USER.md § 分隔）网页搜索 + 抓取双引擎 — 7 引擎 BM25/Reranker + 三层本地抓取 DOM 修剪（$0/月）vs Hermes 云 API 透传 + Firecrawl/LLM web_extractFTS5 + Qdrant 混合搜索 + 作用域/谱系过滤 — Hermes v0.15 仅本地文本（~20ms，无语义检索）108 模式安全扫描 vs Hermes threat_patterns.py（3 作用域 ~20 模式）22+ 中间件流水线 vs 8 层固定栈 — Myrm 各组件可独立配置四层模型纪律 （CORE→ENFORCEMENT→FAMILY→ESCALATION）vs Hermes 模型门控文本块 — Myrm 含 ESCALATION_CONTRACT 自动模型自升级GUI 优先 + Tauri 桌面 vs 仅 CLI（v0.15 加 TUI 多会话，仍终端绑定）4D 预算 （Token + USD + 时间 + 最大子代）— Hermes v0.15 仅每任务超时（1 维）三级智能模型路由 — Hermes v0.15 仅手动每任务选模型（无自动路由）6 层 Prompt Cache vs Hermes system_and_3（4 断点、仅 Anthropic）— Myrm 支持 5+ 提供商 + 缓存断裂检测 + 防抖动
技能进化 — 真·自进化 Agent Myrm 技能进化（42 模块、原生内置）实现 Self-Improving Agent 研究可工程化概念。Hermes「自进化」是第三方 AGPL darwinian_evolver 的外部 CLI 包装，非原生能力。
能力 Hermes Myrm Integration External CLI wrapper (AGPL-3.0) Native built-in (42 modules, MIT)Auto-learn from conversations ✅ Basic ✅ CAPTURED + structured extraction + deduplication Auto-fix failed skills ❌ ✅ FIX + Retrieve-Before-Generate + evidence-driven Multi-variant competition ❌ ✅ 3 parallel variants + LLM-Judge scoring Evolution cost per run 50-500 LLM calls 3 variants (1% of competitor cost)Safety boundary 1-layer scan ✅ 5-layer (anti-loop + sandbox + GUI approval + validator + evolution lock) Quality monitoring ❌ ✅ 3-dimension degradation detection (success rate + P95 latency + 5xx error rate) User frustration detection ❌ ✅ 5 categories, 38 patterns (Chinese + English) Evidence aggregation ❌ ✅ Success/failure case grouping with common error pattern extraction Description optimization ❌ ✅ Auto-refine trigger conditions for better skill matching GUI approval workflow ❌ Auto-execute ✅ Review, modify, approve or reject before applying Per-agent Insights Inbox (background distill → approve in Agent settings)❌ Global auto-apply ✅ Insights tab — nothing applies until you approve; dismiss writes negative exemplar Per-agent skill isolation (no cross-agent habit pollution) ⚠️ Shared pool ✅ agent_id scoping + CoW fork on cross-agent mounts Growth dashboard ❌ ✅ KPI + heatmap + radar chart + evolution timeline Environment fingerprint ❌ ✅ Cross-device safe skill sharing Batch confirmation ❌ ✅ 90% cost reduction for multi-skill evolution Evolution constraints ❌ ✅ Historical error memory prevents repeated mistakes Prebuilt skill upgrade protection ❌ Silent overwrite ✅ Three-way hash — user edits preserved, “Update Available” badge Daily work journal ❌ ✅ 6-source aggregated daily timeline with date navigation
vs ECC (Everything Claude Code) continuous-learning-v2： ECC 在后台蒸馏原子化「本能」YAML 习惯 — 对高级用户强大，但习惯易全局生效且无 per-agent GUI 门控。Myrm 提供相同 思路 （空闲蒸馏 → 技能提案）+ Agent 设置中的 Insights Inbox ：提案保持草稿直至你批准；驳回写入负例以防重复提案；内置 Agent 在 inbox 只读。MCP 方面 Myrm 已提供 GUI 启用前扫描 + 验证 + 运行时 fail-closed （见下文 MCP 安全门）— 领先 Hermes 仅日志流；我们尚未 提供 ECC 的 /aside 分叉聊天、/context-budget 树图或 102-hook 全量静态包（roadmap #6）。技能模块架构 — 八维优势 除进化外，Myrm 技能系统架构在各维度上更成熟：
领域 Hermes Myrm 用户收益 Tool-Conditional Activation Procedural if/else in prompt builder 4-field declarative (requires_tools/fallback_for_tools + tool group variants)Skills auto-adapt to available tools across Web/Tauri/SaaS Skill Injection Manual @skill-name selection Zero-roundtrip auto-injection via slash commands + template variables (${SKILL_DIR}, !cmd“)Say one word, skill activates instantly Config Management Plain text KV injection ([Skill config: key=value]) — secrets visible to LLM JSON Schema standard with auto-generated forms + multi-instance + env_overrides (secrets never reach LLM)Secure API key management + same skill, multiple configs Prebuilt Skill Protection No migration path (pre-hash users lose protection) Three-way hash with GUI “Update Available” badge + accept/reject workflowCustomize freely, never lose your changes Installation Security Hardcoded 4×3 policy matrix Dynamic trust recommendation (26 categories/108 patterns + AST + LLM audit) with 83 security testsIntelligent security that adapts, not static rules Curator Governance Coarse-grained skip (foreground skills never managed → list bloat) Non-destructive archive (always recoverable) + pinned + evolution_lockedSkills stay organized, important ones stay protected Missing Dependency Handling Load full doc then append setup note XML summary shows reason at L0/L1 (save tokens) + 3-level fallback degraded docsKnow instantly what’s missing, fix it fast
从 Hermes 迁移（技能系统） Hermes Skill Feature Myrm Equivalent 体验 /skill-name manual activationSlash command auto-injection ⬆️ Upgrade [Skill config: key=value] configJSON Schema form + env_overrides ⬆️ Upgrade (secrets safe) skills_guard security scan4-level trust + 26-category scan + AST + LLM audit ⬆️ Upgrade INSTALL_POLICY matrixDynamic SkillTrustRecommendation ⬆️ Upgrade lock.json provenanceorigin.json + DB-backed trust + version history ⬆️ Upgrade curator --dry-run CLIGUI Curator panel with preview ⬆️ Upgrade skill_provenance ContextVarpinned + evolution_locked + non-destructive archive ⬆️ Upgrade 176 skills (quantity) 28 precision skills (quality + structured contracts) ↔️ Tradeoff (quality vs quantity)
结论：7 项升级、1 项权衡、0 项降级。用户获得智能安全、安全配置管理与有序技能生命周期，零能力损失。 vs MiniMax Mavis — 多 Agent 团队平台 MiniMax Mavis 是闭源 SaaS 多 Agent 系统，采用 Leader-Worker-Verifier 架构，仅通过飞书（Lark）集成提供。
Mavis 做得好的地方
Leader-Worker-Verifier 模式 — 规划、执行、验证角色清晰分离IM 原生体验 — 飞书聊天内直接多 Agent 协作「即时回复」+ 后台执行 — 立即确认用户，任务异步运行Worker 间上下文隔离 — 各 Worker 独立运行
Myrm 的领先之处 领域 MiniMax Mavis Myrm 用户收益 Orchestration Leader LLM dispatches tasks 6 deterministic modes (Spawn/Chain/Batch/DAG/Verified/Swarm Fission)Structured scheduling, not dependent on LLM improvisation DAG Dependencies ❌ None ✅ Dependency graph + cycle detection + concurrency limits Complex tasks auto-resolve execution order Verification Text-based review Physical evidence enforcement — must provide STDOUT/STDERR execution logs; claiming PASS without evidence is forced to FAILNo rubber-stamp reviews Tool Isolation Not disclosed 5-layer isolation (type admission → global blocklist → per-config → parent-child intersection → role control)Sub-agents cannot escalate privileges Budget Control Token Plan (commercial tier) 4-dimension control (Token + USD + time + max descendants)Precise cost management, not plan-tier anxiety Model Freedom Locked to MiniMax models 100+ models with 3-level intelligent routingZero vendor lock-in Channels Lark only 35+ channels (WeChat/DingTalk/Slack/Telegram/Discord/Email…)Not limited to one platform Deployment Closed-source SaaS only Web + Tauri + SaaS (MIT open source)Full data sovereignty Skill Evolution Basic “learned something” memory update 42-module native system with A/B testing + GUI approval + 5-layer safetyTrue self-improvement vs simple memory Checkpoint Recovery Not mentioned ✅ Checkpointer saves stage-by-stage state Long tasks survive crashes Cost Efficiency Users report “tokens are burning” DelegationBudget + intelligent routing saves 60-80% Zero cost anxiety Auditability Closed-source, not auditable MIT open source + 14 EventKind audit trail Enterprise compliance
从 Mavis 迁移到 Myrm Mavis Feature Myrm Equivalent 体验 Leader-Worker-Verifier DAG + Verification orchestration mode ⬆️ Upgrade (6 modes vs 1) IM multi-task parallel WebUI multi-session + Goal continuation ↔️ Equivalent (different design) Verifier adversarial check _enforce_evidence + ReadonlySandbox⬆️ Upgrade (physical evidence required) Plan → Approve → Execute PlannerAgent + Goal approval workflow ⬆️ Upgrade (7-state lifecycle) “Learned something” 42-module Skill Evolution system ⬆️ Upgrade (full evolution pipeline) Lark integration 35+ channel support (including Lark) ⬆️ Upgrade Worker context isolation CoW workspace_isolation + ArtifactVault ⬆️ Upgrade (engineering-grade) Token display 4D budget + real-time tracking + alerts ⬆️ Upgrade
结论：7 项升级、1 项持平、0 项降级。用户获得开源数据主权、模型自由与多平台接入，零能力损失。 vs Claude Code — Fork 子 Agent 与 Prompt Cache Claude Code 采用「Fork 子 Agent」设计，字节级 prompt 前缀对齐以复用 KV Cache，子 Agent 成本最高可降 90%。
Myrm 的领先之处 领域 Claude Code Myrm 用户收益 Prompt Cache Architecture 1 strategy (prefix alignment) 6-layer system (tool layering → system freeze → explicit breakpoints → cache-friendly compression → pipeline → LLM layer)Comprehensive caching, not just alignment Fork Context Reuse Empty system prompt (loses cache prefix) Full prefix reuse (context_mode="fork") → 100% Prefix Cache HitHigher cache hit rate, lower cost Cache Breakpoints None 4 strategies (after system / every 15 blocks / after compression / last message)Anthropic best-practice coverage Cache Break Detection ❌ None 2-phase attribution with 5 root causes (system/tools/model/TTL/eviction)Quickly diagnose why cache dropped Anti-Thrashing ❌ None ✅ Prevents compression from repeatedly invalidating cache Stable caching in long sessions Resume-Aware Cache ❌ None ✅ Preserves cache prefix on session resume (90% savings in resume scenarios) Cost-efficient session continuity Multi-Provider Cache Anthropic only 5+ providers (Anthropic + Qwen + OpenAI + DeepSeek + Gemini)Zero vendor lock-in Sub-Agent Tool Control All-or-nothing (Fork = no tools) 5-layer isolation (type → blocklist → per-config → parent-child → role)Fine-grained per-tool control Cost Budget None 4-dimension (Token + USD + time + descendants)Precise cost management Observability CLI text NDJSON metrics + Cache Metrics Collector + 5-level monitoring Full-stack cache visibility
从 Claude Code 迁移到 Myrm Claude Code Feature Myrm Equivalent 体验 Fork Subagent (lightweight child) context_mode="fork" with prefix reuse⬆️ Upgrade (100% cache hit vs lost prefix) Prompt Cache prefix alignment 6-layer Prompt Cache system ⬆️ Upgrade (6 layers vs 1 strategy) Agent Team (parallel agents) 6 orchestration modes (Spawn/Chain/Batch/DAG/Verified/Swarm Fission) ⬆️ Upgrade sendTask message passing P2P Mailbox + AgentHandoverState ⬆️ Upgrade (structured handover) CLI interface Web GUI + Tauri Desktop + SaaS ⬆️ Upgrade Anthropic-only models 100+ models with intelligent routing ⬆️ Upgrade
结论：6 项升级、0 项持平、0 项降级。Myrm 缓存经济性更强，具备 Claude Code 完全缺失的三项能力（断裂检测、防抖动、Resume 感知）。 Claude Code 2.1.154~2.1.157 Harness 升级 随 Opus 4.8，Claude Code 推出 /effort（6 级推理预算）、Dynamic Workflows、精简 system prompt 与增强 --resume。Myrm 对比如下：
Feature Claude Code Myrm 结果 Reasoning budget control /effort 6 levels (manual)complexity_router 3-tier auto-routing + PenaltyTracker⬆️ Smarter (auto vs manual) Workflow orchestration Dynamic Workflows (unpredictable LLM JS, keyword trigger) 6 deterministic modes + manual Dynamic Workflow (Python PTC + SQLite event sourcing) ⬆️ Dual-path: safe DW + richer modes System prompt optimization lean system prompt 6-layer Prompt Cache + TaskAdaptiveMiddleware ⬆️ Complete cache engineering Autonomous execution ”Ask fewer questions” Goal 7-state machine + auto_approve + continuation guard ⬆️ Systematic Skill system .claude/skills directory scan 42-module Skill system + 6 discovery sources + evolution ⬆️ Full evolution ecosystem Agent configuration CLI-based agent config 19+ templates + GUI CRUD + DB storage + 5-layer isolation ⬆️ Visual management Code isolation EnterWorktree WorkspacePolicy (INHERIT/ISOLATED_COPY/READ_ONLY) ⬆️ 3 policies Task resumption --resume CLI flagFull checkpoint system + OrphanRecovery + StreamRecovery ⬆️ Complete recovery Cost control Manual effort selection Auto-routing + GoalBudget + DelegationBudget + token_economics ⬆️ Automated Model support Claude-only 100+ models with intelligent routing ⬆️ Model freedom
结论：10 项升级、0 项持平、0 项降级。 动态工作流 — 真实痛点 基于 Claude Code Dynamic Workflows 用户反馈（2026-05 数据）：
User-Reported Issue Root Cause Myrm Solution ”Planned 47 agents, only 25 actually ran” LLM-generated JS scripts are non-deterministic DAG for production; Dynamic Workflow replays completed sub-agents from SQLite cache on retry”Started with 10, ballooned to 82, burned all tokens” No budget enforcement in runtime DelegationBudget — 4D hard limits (Token+USD+time+descendants), impossible to overshoot”8 万字 output lost content midway” No checkpointing for long parallel runs Checkpointer + OrphanRecovery + WorkflowEventStore — stage-by-stage persistence, crash-safe”Had to babysit for 5 hours” No completion guard or auto-recovery CompletionGuard + 429/503 auto-backoff + 7×24 unattended operation”Accidentally triggered workflow by keyword” Keyword-based trigger (mention “parallel” or “research”) Manual WorkflowModeToggle — opt-in only, auto-resets after send; zero keyword accidents”max 16 concurrency, max 1000 total agents” Fixed runtime limits Configurable ConcurrencyLimiter + TokenBucket — no hard ceiling”Cost is a black box until the end” No real-time cost tracking SubagentDashboard — Real-time $0.001 precision cost visualization per node”Approval requests get lost in long terminal logs” CLI-only, no visual anchors Visual HITL Approval — One-click jump from Dashboard to Approval Card”Hard to review what the agent is doing” Raw JSON arguments in terminal Polymorphic Views — Syntax-highlighted diffs for code, terminal UI for shell
关键架构差异 ：Claude Code DW 运行时生成带关键词触发的 JavaScript 编排脚本（非确定性，可失败或偏离）。Myrm 双路径 ：生产负载用声明式 DAG（确定性、执行前可验证）；可选 Dynamic Workflow （手动开关、Python PTC 沙箱、确定性 workflow_id、SQLite 事件溯源实现子 Agent 幂等重放）用于临时并行脚本 — 均配 一等 HITL GUI 。vs Scrapling & BrowserUse — 全自主混合浏览器引擎 典型 AI Agent 用 Playwright/Selenium 包装（如 BrowserUse）在动态页频繁崩溃；传统抓取框架（如 Scrapling）需开发者手写反爬代码。Myrm Agent 提供 全自主混合浏览器引擎 。
Scrapling & BrowserUse 做得好的地方
BrowserUse : Wraps Playwright for LLMs to click elements, but relies on slow LLM reasoning every time a locator fails.Scrapling : Provides powerful stealth tools (camoufox / curl_cffi HTTP fetching) but requires developers to manually wire them into a crawler script.
Myrm 的领先之处 领域 Traditional Agents / Scrapers Myrm Agent 用户收益 Self-Healing Locators & Shadow DOM Piercing Wait 30s for LLM to rethink after DOM changes, completely blind to Shadow DOM O(1) Millisecond Healing Engine Agent instantly recovers from dynamic DOM/React changes using absolute BBox + ARIA implicit mapping + Semantic Veto. Fully pierces Shadow DOM and uses strict W3C CSS whitelists for zero data-pollution. Zero LLM delay. Native Spatial Engine Fails on dynamic layouts, slow Python-level bounding box calculations Direct C++ Layout Selectors Zero-latency, highly accurate element targeting based on visual layout (e.g. “right-of”, “below”) seamlessly native to the rendering engine, with 0 LLM hallucination risk. Render Engine Degradation Heavy Chromium renders everything Autonomous Hybrid Engine Detects static pages and injects HTTP pre-fetch via page.route — skipping JS/CSS load. 90% faster and lighter. Anti-Bot Evasion Fails at Cloudflare (BrowserUse) / Manual config (Scrapling) Dual Stealth Hot-Swap Seamlessly escalates from HTTP → Patchright → Camoufox stealth engine automatically when challenged. Proxy Pool Rotation Global network hooks cause cross-task pollution & lost login state Zero-Network V8 Injection Swaps IPs instantly while injecting local storage via V8 initialization. Perfect state inheritance with exponential backoff. Element Safety Blindly clicks overlapping elements Semantic Veto Prevents mis-clicks using strict contextual checking (even in Chinese/English i18n).
迁移收益 从原生 Playwright 或其他 Agent 框架迁移的用户获得：
Zero-Crash UI Navigation : End the frustration of “Element not found” errors ruining a 20-minute agent task.Blazing Fast Data Retrieval : Don’t burn memory on full Chromium instances for simple text extraction; Myrm degrades to HTTP seamlessly.Production Grade Reliability : Handles CAPTCHAs and Cloudflare invisibly.
凭证库 — 密码与 2FA 永不进入 LLM 多数框架在自动化登录时的默认模式是灾难性的：模型生成密码字符串并通过 type/fill 工具参数传递，该值残留在聊天历史、MCP 日志与重试缓冲区。
Myrm 表单凭证库 将「知道凭证」与「使用凭证」分离：
Layer What happens You Configure labels in Settings → Credentials (password + optional TOTP seed) Agent / LLM Sees label names only; calls fill_credential (browser) or type_credential (desktop) Harness Decrypts in memory, injects at DOM or OS layer — plaintext never returns to context
Myrm 的领先之处 领域 Hermes / OpenClaw FSB (Chrome extension) Myrm 用户收益 Vault boundary ❌ Plaintext in tool args ✅ Browser DOM injection ✅ Browser + desktop Login automation without password in chat TOTP / 2FA ❌ Manual ❌ Not documented ✅ Built-in RFC 6238 Agent completes 2FA without you reading codes Management UI ❌ Env vars / chat Extension popup ✅ WebUI Settings panel One place for all automation credentials Security stack integration ⚠️ Post-hoc patches Standalone extension ✅ 6-layer defense + leak detection + audit Vault is part of enterprise security, not a bolt-on
与 FSB 的诚实对比 FSB 首创浏览器自动化的库边界模式（标签引用 → 扩展解密 → DOM 填充）。Myrm 采用 相同安全原则 ，并扩展到 桌面 Computer Use 、原生 TOTP 与 统一产品 GUI — 无需独立 Chrome 扩展。FSB 在支付卡 API（use_payment_method）仍领先；Myrm 今日覆盖密码字段与 2FA。
迁移收益
从 Hermes / OpenClaw ：别再往提示词里贴密码；在设置中配置一次即可。从 FSB ：相同心智模型（标签），外加桌面应用与 TOTP，同一工作区。企业场景 ：凭证库 + 12 维权限 + Merkle 审计（记录调用、不记密钥）+ 无痕会话跑敏感任务。
MCP 安全门 — 启用前先知风险 第三方 MCP 服务器攻击面日益扩大：投毒工具描述、敏感路径访问、运行时工具注入可在对话中途危及 Agent。
Myrm 在 MCP 进入工作区之前 设门：
Stage What Myrm does What you see Typical competitors Edit Debounced static scan (~0.03ms) Amber risk list (EN/ZH) No GUI pre-check Save / enable High-risk requires ack + 4-step verify (static → OSV → connect → runtime) Clear block or confirmed enable Hermes: log only Runtime Harness fail-closed disconnect Poisoned MCP never joins the chat OpenClaw: env filter only
诚实范围 ：完整 102-hook 静态规则包在 roadmap — 本门控覆盖真实用户路径（设置 → 启用 → 聊天）。回归：harness + server API 测试（21+ 用例）。迁移收益
从 Hermes ：别只在日志里发现坏 MCP — 先在设置中拦截或确认。从 OpenClaw ：环境过滤不够；要启用前扫描 + 运行时断开。
Shell 命令可视化审批 — 允许前看清每条管道 Agent 执行 curl … | bash 时，单行等宽字体隐藏了哪段下载、哪段执行。
Myrm Shell 命令展示 将管道拆为带逐段风险着色的 span — 与 OpenClaw 用户心智模型一致，默认开启并接入六层安全栈。
领域 OpenClaw Hermes Myrm User benefit Pipe breakdown tree-sitter explainer (not on by default) Whole-line allow/deny On by default , same source as redacted displayKnow which segment is dangerous Risk coloring ❌ ❌ ✅ Per-span levels Red/yellow/green at a glance Secrets in UI Basic Basic Redact first, then spans API keys never in approval text Edit escalation ❌ ❌ Block edits that change UNKNOWN-risk commands Can’t “tweak wording” into a worse command Sub-agent cards ❌ ❌ ✅ Spans flow to delegate approvals Same clarity for spawned workers Workspace context ❌ ❌ ✅ Shows workspace root (EN/ZH) Know where the shell runs
诚实限制 ：超过 128KB 的命令截断解析（UI 提示）。生产 PyPI 安装需 harness 发版；本地 ./myrm dev 用 editable harness（45 harness + 33 前端审批测试）。迁移收益
从 OpenClaw ：熟悉的分段 shell 视图 — 外加 12 维权限、泄漏检测与 Merkle 审计。从 Hermes / Claude Code ：别盲批单行命令 — 一键前看清管道段与风险。
vs MemPalace — AI 记忆系统（14.9K+ Stars） MemPalace 是独立 AI 记忆系统，用 Wing/Room/Closet/Drawer 建筑隐喻，奉行「逐字存储一切」，LongMemEval 达 96.6% R@5。作为 MCP 工具供外部 AI 助手调用。
MemPalace 做得好的地方
Verbatim storage — stores raw conversations without lossy summarizationArchitectural organization — Wing/Room/Closet/Drawer hierarchy gives AI a “navigation map”4-layer memory stack — L0 identity (~50 tokens) through L3 deep search, keeping wake-up cost under 900 tokensMulti-format ingestion — normalizes Claude, ChatGPT, Codex, Gemini, Slack exports into a unified formatLocal-first — runs entirely on your machine with ChromaDB, zero cloud API calls
Myrm 的领先之处 领域 MemPalace Myrm 用户收益 Memory Types 1 flat type (drawer) 8 structured types (Profile/Semantic/Episodic/Procedural/Conversation/Claim/TaskDigest/Integration)AI understands “this is a fact” vs “this is a preference” vs “this is a rule” Verbatim Storage ChromaDB single vector Dual-track (raw_exchange + summary) with dual embedding Both precise recall and fast overview Deduplication Single-layer vector similarity 3-layer smart dedup (Hash → Vector → LLM) with UPDATE_REPLACE/UPDATE_MERGE/NEWNo duplicate memories, intelligent merging Forgetting None (memories only grow) 5-dimension intelligent forgetting (time decay + access frequency + importance + relations + user rating)Memory stays lean and relevant Knowledge Graph Basic halls/tunnels co-occurrence GraphStore + CTE with visual explorationRich relationship mapping Contradiction Detection Rule-based name/relationship check LLM-powered ClaimGraph + cognitive subsumptionCatches subtle contradictions Search BM25 + Vector hybrid FTS5 + Qdrant hybrid (FTS5 has built-in BM25)Production-grade search infrastructure GUI Management ❌ CLI only ✅ Full GUI panel — browse, edit, approve, pin, delete memories Visual memory management Memory Safety ❌ None ✅ Scanner + sanitizer with real-time security events Prevent memory poisoning Health Diagnostics ❌ None ✅ Benchmark testing (Recall@K, NDCG, MRR, Precision) Quantified memory quality Preference Tracking ❌ None ✅ Stability scoring with visual trend cards Track how preferences evolve Backup & Restore Manual SQLite copy ✅ Structured backup/restore protocol Data safety built-in Platform Memory library only Complete AI Agent platform (100+ tools, 35+ channels, Sub-Agent, Goals)Memory is part of a full workspace Data Storage ChromaDB (known HNSW corruption issues) Qdrant + SQLite (production-grade)No manual HNSW repair needed
从 MemPalace 迁移到 Myrm MemPalace Feature Myrm Equivalent 体验 Verbatim drawer storage ConversationMemory.raw_exchange (dual-track) ⬆️ Upgrade (raw + summary) Wing/Room/Closet hierarchy 8 memory types + GraphStore knowledge graph ⬆️ Upgrade (typed + relational) 4-layer memory stack (L0–L3) 5-layer context pipeline + on-demand retrieval ⬆️ Upgrade ChromaDB vector search Qdrant dual-vector + FTS5 hybrid ⬆️ Upgrade (more robust) MCP tool integration Native MCP server + agent-integrated tools ⬆️ Upgrade (native, not add-on) Local-only operation Local + Tauri desktop + SaaS (your choice) ⬆️ Upgrade (3 deployment modes) CLI management GUI panel with visual memory management ⬆️ Upgrade mempalace mine data ingestionReal-time auto_extract + tool_capture ⬆️ Upgrade (no manual mining) mempalace searchAgent-integrated memory recall with source citations ⬆️ Upgrade Hallway connections Knowledge graph with CTE traversal ⬆️ Upgrade
结论：10 项升级、0 项持平、0 项降级。MemPalace 用户获得记忆原生集成的完整 Agent 平台（非外挂），并多出 12 项 MemPalace 不具备的能力（智能遗忘、GUI 管理、安全扫描、偏好追踪、健康诊断等）。 vs Claude Office Visualizer — CLI 状态仪表盘 Claude Office Visualizer 用像素风办公室动画渲染 Claude Code CLI 状态，通过独立 Next.js + PixiJS 应用展示 Agent 状态、上下文用量与任务进度。解决 CLI 用户不盯终端就看不见 Agent 在干啥的痛点。
Claude Office 做得好的地方
Visual agent metaphor — pixel characters that walk, think, and interact based on real CLI state12 whiteboard modes — multiple visualization layouts for different data viewsTour overlay — 7-step interactive guide for new usersAttention system — screen flash notifications when the agent needs user inputDocker deployment — easy self-hosted setup
为何 Myrm 不需要这个 Myrm 是 GUI 优先 应用。Claude Office Visualizer 要解决的「看不见 Agent 状态」「CLI 输出无聊」— 在 Myrm 架构中 不存在 。
领域 Claude Office Visualizer Myrm 用户收益 Agent Status Pixel animation in separate window SubagentDashboard — structured panel with real-time progress, inlinePrecise data, no extra windows Context Usage TrashCanSprite (filling trash can metaphor) ContextUsageIndicator — exact percentage + multi-level color warningsNumbers > metaphors Task Management 12 PixiJS read-only whiteboard modes KanbanBoardView — interactive drag/drop/filter/editActionable, not just viewable Notifications Screen flash (attentionStore) NotificationBell — 4 severity levels + unread count + persistenceStructured alerts, not screen flashing Onboarding TourOverlay 7-step walkthrough EmptyChat + SamplePrompts + CompetitorMigrationBanner Conversation-native guidance Companion Fixed pixel character 15 species + 9 hats + 5D attributes + evolution + shiny variants Full RPG companion system Deployment Docker front+back separate deployment Built into the app — zero extra infrastructureZero additional setup Mobile ❌ Desktop-only (PixiJS) ✅ PWA + 35+ messaging channels Monitor from anywhere Bundle Cost +~200KB (PixiJS engine) 0KB extra (SVG vector rendering)No performance penalty
从 Claude Office Visualizer 迁移 Claude Office Feature Myrm Equivalent 体验 Pixel agent animation CompanionSprite (15 species + evolution) ⬆️ Upgrade 12-mode whiteboard KanbanBoardView + GoalDagRenderer + EventTimeline + GrowthDashboard ⬆️ Upgrade (interactive) TrashCan context display ContextUsageIndicator (precise metrics) ⬆️ Upgrade attentionStore notifications NotificationBell (4-level, persistent) ⬆️ Upgrade Tour guide EmptyChat + SamplePrompts ↔️ Equivalent (different product UX) Docker deployment Tauri/Web/SaaS (3 deployment modes) ⬆️ Upgrade
结论：5 项升级、1 项持平、0 项降级。用户从独立观察窗口进入数据精准、可交互、完整 Agent 平台的一体化 GUI。 vs Coze 3.0 / Lobster / Vercel v0 — 产物部署与只读链接（Sites 2.0） Coze 将项目锁在生态内；Lobster 擅长公开静态链但非从 Agent 工作区部署 Vercel；v0 面向用 AI 开发的开发者。Myrm 面向 希望从聊天产物获得可分享结果的 GUI 用户 ，非独立建站工具。
维度 Coze 3.0 Lobster Vercel v0 Myrm Entry point Stay in Coze Export/share flow Dev-oriented generator Deploy or Link from artifact preview Formal deploy Platform-hosted Not core Vercel-native One-click Vercel + preflight gateRead-only link Platform-dependent Public link (often gated) N/A for chat artifacts 7-day signed Link , no Vercel requiredMulti-file HTML Varies Strong directory share Often single-page Same bundle as deploy ; trailing-slash for css/jsLock-in Closed hosting Subscription walls on some tiers Vercel account Open — Local BYOK or sandbox platform token
你将获得 ：聊天产物 → 落地页/报告/迷你应用 → ~30s 部署 URL 或 即时只读 Link 供审阅。诚实限制 ：尚无分享撤销/分享码 UI；长期公开站点应用 Deploy + 自有域名；纯 code 产物须先导出为 html（预检 + UI 门控一致）。vs Codex (OpenAI) — Appshots 与 /goal GA Codex 近期发布两大旗舰功能：Appshots （⌘⌘ 窗口捕获 + 文本提取）与 /goal mode GA （长时间自主任务）。Myrm 对比如下：
Appshots（窗口捕获 + 文本提取） 维度 Codex Myrm Platform ⚠️ macOS only ✅ macOS + Windows + Linux Text extraction Window text incl. off-screen window_text() via AX API / uiautomation / xdotoolDPI handling Not mentioned Binary-search downsampling prevents Retina coordinate drift Return format Screenshot + text list[ContentBlock] structured multimodal (text + image)Safety Computer Use permission 5 blocked key combos + dangerous-text regex + TOCTOU revalidation after approval delay + GUI BBox approval card Visual approval Text-only or iPhone push Inline BBox + AttentionBar + Tauri OS red frame on the target monitor (screen-absolute coords, multi-monitor match)
你将获得 ：Agent 要点桌面时，你能看到 位置 — 不仅是文字「批准？」。Tauri 上 系统级红框 在聊天窗口被挡时仍可见。诚实限制 ：OS 叠加层冒烟需 Tauri 桌面端；仅浏览器 WebUI 有内联 + AttentionBar，无 OS 框。/goal 模式 维度 Codex /goal Myrm Goal System Planning Linear self-planning PlannerAgent auto-generates DAG plans with dependencies Execution Linear, sequential DAG concurrent executor + Swarm Fission Budget No budget control 4D budget (tokens / USD / wall-clock / turns) Completion User checks CompletionGuard — evidence-based TDD-like verification Continuation ”Asks if stuck” 7-step guard chain + Semantic Judge Multi-goal Single goal Priority Queue + auto_approve unattended serial Runtime adjust Not mentioned Dynamic Subgoals + Objective Hot-Edit Git awareness Not mentioned Branch-Aware Stash & Migrate Frontend Check progress GoalControlPlane real-time panel + Execution Summary card Notification Built-in channel_notify_tool — Telegram/Slack/any channel
锁屏与远程 维度 Codex Myrm Background run Mac lock screen (not lid close) SaaS: naturally unaffected; Local: GracefulShutdownManager + checkpoint Mobile access ChatGPT App Web frontend reconnects anytime + SSE real-time push
用户痛点（评论反馈）Myrm 已解决 Pain Point Codex Status Myrm Solution Windows not supported ❌ ✅ WindowsBackend with uiautomation Token consumption too fast No budget GoalBudget 4D + 6-layer Prompt Cache Context compression disappeared User complaint 22+ middleware context pipeline Ran 30h nonstop, no auto-stop No limit max_time / max_turns auto-pause Linux partial support ⚠️ dmg-based ✅ LinuxBackend native implementation
结论：Codex 的 Appshots 与 /goal 是 Myrm 现有能力的简化子集。Myrm 覆盖 3 平台（vs 仅 Mac）、DAG 执行（vs 线性），并提供 Codex 完全缺失的企业级预算控制与完成验证。 vs 360 LobsterAI — 消费级 Agent 平台 360 LobsterAI 是消费级 Agent 产品（周鸿祎背书），主打手动模型档位、100+ 预设「专家龙虾」与「教练式」上手流程。
成本智能 维度 360 LobsterAI Myrm 用户收益 Model selection Manual 3 tiers (Lite/Save/Full) Auto ComplexityRouter (SIMPLE/STANDARD/REASONING)No guesswork — system picks optimal model Routing algorithm None (human judgment) 2-phase (rule matching + LLM Judge) Accurate even for ambiguous tasks Session continuity Context lost on tier switch Session Momentum prevents downgradesMulti-turn complex tasks stay on-tier Accuracy feedback None PenaltyTracker — learns from misroutesGets smarter over time Privacy routing None PrivacyRouter — sensitive data stays localData sovereignty built-in Budget control None 3D×3-level auto budgeting + SSE alerts Never overspend Cost visibility Basic stats 15+ dimension dashboard + cache economics Full cost transparency per message
Agent 模板 维度 360 LobsterAI Myrm Preset agents 100+ (quantity) 19 high-quality + unlimited custom + Skill Marketplace Template depth System prompt only System prompt + tools + skills + model + security overrides Customization Limited editing Full GUI editor + clone + import/export + version snapshots Onboarding Q&A “coaching” wizard PresetAgent Gallery + conversational creation + per-agent suggestion prompts Anti-confusion None 5-layer anti-interference (Profile → Conditional Activation → Progressive Disclosure → Noise Gauge → Hybrid Retrieval)
多渠道接入 维度 360 LobsterAI Myrm Channels 3-4 (Feishu, DingTalk, App) 35+ providers (Telegram, Slack, Discord, WeChat, WhatsApp, email, and more)Per-channel binding Not mentioned Each channel/topic can bind a different agent
从 360 LobsterAI 迁移到 Myrm 从 360 LobsterAI 迁移的用户获得：
Automatic cost optimization — no need to manually select “lite” vs “full” modeSmarter routing — system learns preferences and improves over time35+ channels vs 3-4, with per-channel agent bindingEnterprise security — 6-layer defense vs consumer-gradeUnlimited extensibility — Skill Marketplace + custom agents vs fixed presetsFull data ownership — self-hosted option, no vendor lock-in
智能多平台分发与 OTA（vs Hermes / Cursor） 分发桌面 Agent 时，传统竞品迫使用户分辨 CPU 架构（Apple Silicon vs Intel），或 CI/CD 覆盖 release manifest 导致静默更新失败。
Myrm 提供 零摩擦分发与 Map-Reduce OTA 更新 ：
领域 Hermes / Cursor (Early) Myrm 用户收益 Mac M-Series Downloads Requires manual user selection (ARM vs Intel) or accidentally delivers Rosetta-bound Intel binaries on Safari due to UA spoofing. Hardware WebGL Sniffing Pierces Safari’s “Intel Mac OS X” UA mask by reading the unmasked GPU renderer. 100% accurate, zero-click correct DMG delivery. iPadOS Spoofing Serves useless .dmg installers to iPads because Safari claims to be macOS. Physical Touchpoint Interception Uses maxTouchPoints to detect iPad hardware, gracefully redirecting users to the SaaS Cloud WebUI. OTA Auto-Updates (CI/CD) Multi-platform CI jobs race to overwrite latest.json, breaking auto-updates for users on the slower-building OS. Map-Reduce CI Architecture Platforms build in isolation, artifact their manifests, and a final cloud job merges all latest.json objects securely. No race conditions. Release Confidence ”Upload and hope” Automated sanity checks on the merged manifest before GitHub Release Users never get stranded on broken auto-update channels.
结论：迁移用户获得首日零摩擦安装体验与坚如磐石的后台更新。 本地迁移向导（v1.4） Myrm 在 Local 与 Tauri 部署提供 GUI 优先竞品迁移向导 ，范围诚实：五个文件系统发现源、四条预览通道，以及 Hermes CLI 可能自动接线但需手动的集成项。
五个发现源（Local / Tauri） Source What imports automatically What you configure manually Hermes Persona, global memory, skills (review queue) Some MCP/channel extensions OpenClaw Memory MD, multi-workspace merge, sessions.json episodic Channels, gateway keys Claude Code Skills, persona; instruction-only dry-run when memory lane empty MCP servers in Settings Cursor Rules, skills, memory fragments IDE-specific paths vary by OS Codex Config and memory exports where present Plus subscription not migrated
四条预览通道
Persona → Agent — SOUL-style instructions attach to a target agent profile.事实 → 全局记忆 — 结构化记忆行，支持批量 回滚 。Skills → Review queue — imported skills require approval before activation; optional Agent binding .API keys → Opt-in — never silently copied; user confirms each secret.
工作流 Scan → Preview (dry-run) → Confirm → Result in Settings → Migration. Server-side resolve_competitor_import_source() forces the correct adapter (fixes OpenClaw source=auto mis-routing to Hermes).GUI 优先无缝技能迁移与原子持久化 用户导入第三方生态技能包（如 Hermes ZIP）时，竞品常依赖基础文件覆盖（os.replace）与顺序数据库写入。SaaS 高并发下会导致灾难性「分裂脑」（DB 有记录但文件缺失）或 API 冻结。
Myrm 提供 零成本热迁移与原子持久化引擎 ：
领域 Hermes (Competitor) Myrm 用户收益 Import Security Basic guard checks Pre-emptive AST interception + Persistent Claim-Check Staging Completely prevents Zip Bomb, Zip Slip, and Memory OOM attacks during drag-and-drop. Metadata Fidelity Brutal overwrite YAML Deep Merge Engine 100% losslessly preserves version, author, and custom extensions in third-party Frontmatter. Database Transactions Sequential loop writes SQLite Executemany Bulk Transaction (BEGIN…COMMIT)All-or-Nothing atomicity — zero risk of phantom skill records. Directory Split-Brain In-place file replacement Blue-Green Swap + Atomic Rename Skills update atomically via .tmp → .old swapping, ensuring directories are never half-written. UI Freezing Synchronous GC blocks main loop FastAPI BackgroundTasks Asynchronous GC Deletes expired staging files and old directories in the background, keeping API latency at 0ms.
结论：50 用户同时上传技能 ZIP 时竞品服务易崩溃或泄漏线程。Myrm 大规模并发导入 0ms 阻塞主线程、零脏写风险。 vs Hermes claw migrate 维度 Hermes CLI migrate Myrm Wizard Interface Terminal GUI with lane-level preview Rollback Limited Memory import batches rollback Skill governance Copy folders Review queue + Agent binding Breadth Broader MCP auto hints Narrower but explicit manual lanes for MCP/channels
诚实评分 ：本地五源 GUI 迁移 9.7/10 — 非「覆盖每个竞品工件」。SaaS 云沙箱不可用 （无法访问用户主机文件系统）；请用 Local WebUI 或 Tauri 桌面端。迁移后你将获得（面向用户）
保留人格与习惯 — SOUL 式指令落在你选的 Agent 配置，非一刀切默认。保留结构化记忆 — 事实与会话导入支持批量回滚。技能可控 — 导入技能进审核队列，你批准并绑定到正确 Agent。无静默偷 Key — API Key 仅在你逐通道 opt-in 时导入。诚实后续 — MCP 与消息渠道在设置中引导（不宣称与 Hermes CLI 提示一键渠道对等）。
产品分层 ：Local/Tauri 上开源 myrm-agent-frontend （向导 UI）+ myrm-agent-server （发现/预演/确认 API）；闭源 harness 仅提供导入类型契约 — 无独立迁移 UI。选择 Myrm 你将获得
永不丢上下文 42+ 模块记忆系统 — 业内最完整的 AMO 实现。8 类记忆、7 信号检索融合、7 层陈旧防御、跨会话整合与一键回滚。Anthropic Dreaming 与 Mem0 部分覆盖的能力，Myrm 端到端交付。
随处工作 通过 35+ 消息渠道，在任何设备上审批任务、引导 Agent、监控进度。
智能工具使用 三层工具分层 + 按需加载 + 三维健康监控 + 14 类错误诊断与专家修复建议。工具精准选用、高效执行、失败自愈。
企业级安全 六层纵深防御：预算控制、PII 保护、污点追踪与审计追踪。
零厂商锁定 100+ 模型，自托管或云端。数据始终归你。
vs OpenClaw 2026.6.1 — 跨越”能用”到”好用”的终极形态 OpenClaw 2026.6.1 是其重磅更新，主打 Windows 原生节点、技能工坊、工作看板以及稳定性修复。然而，从其官方发布和大量用户社区反馈来看，其依然面临着”更新即崩”（如飞书/QQ断连）、任务缺乏直观可视化干预、以及”缝合感”较强的痛点。
Myrm 在架构和体验设计上实现了降维打击：
Myrm 的领先之处 维度 OpenClaw 2026.6.1 Myrm 用户收益 (Migration Win) 桌面原生体验 仅提供 Windows Companion 节点环境 Tauri 桌面端 + macOS 刘海屏状态胶囊 提供系统级沉淀。当任务后台运行时，无需打开窗口，直接看菜单栏胶囊即可了解进度。 任务可视化干预 Workboard 纯展示层 动态看板 + 实时断路器可视化 (Circuit Breaker) 不仅能看，当子任务死循环时，系统自动熔断标红，或在手机端推送交互卡片，用户点击【Unblock】瞬间接管。 渠道通信自愈 ”修复”但评论区频发飞书/QQ掉线 断点续传 + SQLite 离线消息队列 + 指数退避重连 即便网络波动或 Token 过期，长文本也能断点续传，不丢消息不卡死。 技能安全演进 基础 Proposal 流程 GUI 审批流 + Quarantine 检疫隔离 + 静态扫描 AI 学习的新技能必须经过防投毒扫描，用户在前端审批后才能上线，杜绝大模型恶搞。 系统稳定性基建 openclaw doctor 命令行诊断前端一键体检 + OOM 静默防护 (Cgroup-Aware) + mtime 状态存储 在 Docker 或 SaaS 下不再无故消失，通过前端即可了解所有网络、端口和 API 健康度。 交互 UX 细节 命令行打字感 长文自动目录 (TOC) + IME 候选词保护 + Quick/Reasoning 模型秒切 极具质感的现代聊天界面。报告自动生成目录并跟随滚动，聊天流畅度比肩原生 App。
一句话总结 ：从 OpenClaw 迁移到 Myrm，你将告别”修一天跑一小时”的极客折腾期，获得一个拥有全天候健壮队列、极致 UI/UX 细节以及跨端原生体验的企业级 AI 工作站。vs DeerFlow — 彻底告别大模型崩溃死循环 DeerFlow 作为一个优秀的开源智能体框架，在健壮性处理上遇到了典型的业界难题：大模型极易因为长日志和异常输出引发死循环崩溃。Myrm 将 DeerFlow 暴露出的核心痛点转化为原生中间件级别的降维打击解决方案 。
无论是使用 LangChain 还是原生 API，执行长线任务时极易遭遇致命的崩溃死循环。Myrm 实现了免人工干预的 100% 自动容错 ：
能力 Myrm Agent DeerFlow / 普通 LangChain Agent 优势 工具输出预算控制 ✅ 自动截断与落盘 (Spillover) ❌ 仅配置最大字符数，超长直接丢失 永远不会 OOM，且数据不丢 超长数据处理 ✅ 预览摘要 + 本地持久化引用 ❌ 粗暴截断，无法查阅全文 大模型既知道结果超长，又知道去哪里读完整日志 悬空工具调用自愈 ✅ Dangling Tool Call Healer 中间件 ❌ 报错崩溃死循环 自动补齐合成的错误回执，使得消息序列合法，让模型自行修复残缺调用
用户感知收益 ：当别的 Agent 跑到一半因为“日志太长”或“JSON 残缺”而崩溃挂起时，Myrm 会自动截取超长日志存入本地文件，并用一段礼貌的提示语告诉大模型：“输出过长已落盘，请使用读文件工具查看某某路径”。它永远能优雅地处理各种边界崩溃。智能记忆检索（Context-Aware TF-IDF Memory Retrieval） 在注入长期记忆时，普通框架会盲目按时间或事实置信度排序，导致上下文被大量“高置信度但与当前任务完全无关”的噪音记忆占满，反而忘记了核心指令。
Myrm 引入了 TF-IDF 与置信度双重加权检索 ：在注水前，先利用 tiktoken 提取当前对话最近几轮作为 current_context，计算所有记忆与该上下文的余弦相似度，并结合事实的置信度进行双端加权评分（similarity*0.6 + confidence*0.4）。
用户感知收益 ：如果你积累了数百条记忆，迁移到 Myrm 后绝不会体验到因为“记得太多而变笨”的痛点。我们在极低的 Token 预算内提供“最精准”的记忆召回。智能防泄漏与零卡顿的标题生成 (O(1) Anti-Blocking Title Generation) 当用户粘贴超大日志或代码块时，后台生成标题的正则清洗往往会卡死整个服务器（Event Loop Blocking）；此外，廉价模型生成的标题经常带有 Title: 等废话前缀，且容易暴露 API Key 等敏感信息。
Myrm 在底层实现了 O(1) 早期截断 与军工级脱敏管线：
能力 Myrm Agent DeerFlow / 传统框架 优势 防阻塞性能 ✅ O(1) 早期截断 ❌ O(N) 全文正则，易卡死 无论粘贴多大文件，前端 UI 永远丝滑零卡顿 凭证脱敏 ✅ 调用 redact_leaks 抹除 API Key ❌ 原样发送给 LLM 彻底封堵隐式 Token 泄漏与提示词注入漏洞 深度降噪 ✅ 正则剥离代码块、URL、HTML ❌ 原始文本直接生成 避免大模型被无关代码迷惑，标题更精准 智能容错 ✅ 空文本 0 Token 拦截，统一英文兜底 ❌ 仍调用 LLM 或报错 极端输入下不浪费 Token，UI 保持整洁 多语言自适应 ✅ <user_input> XML 隔离强制语种跟随 ❌ 易出现中英混杂 全球化体验更佳
用户感知收益 ：哪怕你粘贴了 1MB 的错误日志，Myrm 也能瞬间为你生成一个安全、精准、无废话前缀的标题，且整个过程对服务器性能零冲击。细节体验 100% 碾压竞品。长文阅读（自动目录） 当 Agent 回复含多个标题时，Myrm 会在消息旁 自动生成目录 — 点击章节跳转，滚动时高亮跟随。ChatGPT、Hermes、OpenClaw 均无等价的聊天内多章节导航；用户通常需复制到 Notion 或 Word 才能获得大纲。
Myrm Typical chat UIs Auto TOC ✅ ≥2 headings ❌ scroll only Scroll sync ✅ ❌ Works while streaming ✅ (debounced) N/A
多 Agent 编排与零信任验证（vs Hermes Agent / Claude Code） 在多 Agent 架构领域，多数框架依赖盲信与非确定性 LLM 即兴发挥。Myrm 打破串行瓶颈，彻底消除 Hermes、Claude Code 等竞品中的「幻觉式成功」与「预算烧穿」痛点。
Myrm 的领先之处 领域 Hermes / Claude Code Myrm 用户收益 Verification Blind trust in LLM summaries Zero-Trust Verification Enforces physical verifiable credentials (e.g., file paths, exit codes). Sub-agents cannot fake success, effectively ending task hallucination. Sandbox Security Background agents can trigger UI UI Interaction Blacklisting LEAF node sub-agents are physically restricted from accessing UI/Chat interfaces, preventing them from unexpectedly interrupting the user.Language Context Sub-agents default to English Dynamic Spec Injection The parent orchestrator dynamically injects the user’s active language and style guidelines, ensuring the final report is completely cohesive (no more “Chinglish”). Budget Control Timeouts only (burns tokens) Iteration Circuit Breaker & Downgrade Introduces a hard limit on iteration steps to stop deadlocks instantly. Automatically downgrades non-core tasks to cheaper models when budgets run low. Human-in-the-Loop Simple “Yes/No” approval Seamless Context Injection During a UI breakpoint approval, users can inject corrective text directly into the agent’s state machine, steering wandering agents back on track with a single click.
迁移收益 从 Hermes 或 Claude Code Dynamic Workflows 迁移，你将获得：
绝对成本可预测 ：Agent 死循环不会让你醒来面对巨额 API 账单。真实执行证据 ：Myrm 称任务完成时，有 OS 级执行日志支撑，非 LLM 想象。无瑕 UI 体验 ：后台 Agent 严格留后台；人工介入可注入纠正语境，非二元是/否。
准备开始？ 
