我阅读了 Hermes Agent 的记忆系统，它修复了 OpenClaw 出错的地方

[

](https://x.com/manthanguptaa)

[

](https://x.com/manthanguptaa/article/2034849672985288957/media/2034849440356544512)

I Read Hermes Agent's Memory System, and It Fixes What OpenClaw Got Wrong

我阅读了 Hermes Agent 的记忆系统，它修复了 OpenClaw 出错的地方

Hermes Agent was particularly interesting to me because this time, I did not have to reverse engineer everything from behavior alone. Hermes is open source, and both the and the are public. So instead of poking a black box with prompts, I went straight to the code paths that build prompt state, persist sessions, flush memories, and query past conversations.

Hermes Agent 特别吸引我，因为这次，我不必仅从行为逆向工程一切。Hermes 是开源的，而且 both the and the 都是公开的。因此，我没有用提示戳黑盒子，而是直接查看构建提示状态、持久化会话、刷新记忆和查询过去对话的代码路径。

The short version is this: Hermes does not have one memory system. It has four.

简短版本是：Hermes 没有一个记忆系统。它有四个。

1. A very small, curated prompt memory stored in `MEMORY.md` and `USER.md`.

1. 一个非常小的、精选的提示记忆，存储在 `MEMORY.md` 和 `USER.md` 中。

2. A searchable SQLite archive of past sessions exposed through `session_search`.

2. 一个可搜索的 SQLite 过去会话存档，通过 `session_search` 暴露。

3. Agent-managed skills that act like procedural memory

3. 代理管理的技能，像程序记忆一样运作

4. An optional layer for deeper user modeling

4. 一个用于更深入用户建模的可选层

And the key design choice tying all of this together is simple: keep the prompt stable for caching, and push everything else to tools.

将这一切联系在一起的关键设计选择很简单：保持提示稳定以便缓存，将其他一切推送到工具中。

Before understanding memory, it helps to understand what Hermes is actually sending to the model.

在理解记忆之前，有助于理解 Hermes 实际发送给模型的内容。

The system prompt is assembled roughly like this:

系统提示大致是这样组装的：

[0] Default agent identity
[1] Tool-aware behavior guidance
[2] Honcho integration block (optional)
[3] Optional system message
[4] Frozen MEMORY.md snapshot
[5] Frozen USER.md snapshot
[6] Skills index
[7] Context files (AGENTS.md, SOUL.md, .cursorrules, .cursor/rules/*.mdc)
[8] Date/time + platform hints
[9] Conversation history
[...