openmem
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
No arguments | |||
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": true
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| mh_writeC | 写入一条记忆/知识到老大的统一记忆中枢（openmem）。source 必填=写入方 agent 名。layer: k=知识 m=记忆。category: rules/facts/projects/lessons/knowledge/archive/verification/services/capabilities/misc（或项目名如 dsh / agents-to-feishu）。重要结论不写入=任务不算完成。 |
| mh_updateA | 按 id 原地更新一条记忆（改错字、补内容、转 pinned、换 category 都用这个）。只改传了的字段，未传的保持不动；改了 content 会自动重算向量与 hash。保留 id，不打断 superseded_by / 冲突记录引用链——优于"删了重写"。改内容前建议先 mh_get 看一眼原文。 |
| mh_searchA | 检索老大的统一记忆，返回原始条目（适合查历史细节、具体事件、某个坑的完整经过）。混合检索：向量语义 + 元数据过滤(layer/category/source/tags/时间)。⚠️ 但「老大是什么人 / 什么偏好 / 本机环境 / 服务与端口 / 某项目怎么改」这类标准问题请先试 mh_tools_list → mh_tool（成品答案，秒回、后台已保鲜），不要直接搜、更不要猜。注意：nssm 服务台账（agent-matrix/services）默认不返回，要查某服务的启动参数/端口用 source="agent-matrix" 显式指定。 |
| mh_serviceA | 查本机 nssm 服务台账：按服务名返回该服务的启动参数/端口/路径等配置明细（不走语义检索，按名精确匹配）。不知道确切名字就先不传 name，会列出全部服务名。 |
| mh_getA | 按 id 取单条记忆完整内容。 |
| mh_askA | AI 对 AI 咨询：像直接问老大陈丹本人一样提问，openmem 基于全部记忆给出完整、准确、口语化的答案（走 litellm GwV4F，数秒级，较慢）。只在 mh_tool 里没有对应成品答案、且问题需要理解+综合时才用：要标准答案走 mh_tool（秒回），要原始记忆片段走 mh_search（更快）。 |
| mh_toolA | 【首选 · 秒回】调用预生成答案工具：标准化提示词 + 后台已备好最新最准的答案，不走 LLM、几乎零等待。凡「老大是什么人 / 偏好习惯 / 铁律清单 / 本机环境 / 服务与端口 / 12 bot 花名册 / 项目索引 / openmem 使用手册」这类标准问题一律先用这个，别用 mh_search 现搜、更别凭印象猜。不知道有哪些成品答案就先调 mh_tools_list（一次看清全部 + 新鲜度）。 |
| mh_tools_listB | 列出全部预生成答案工具及答案新鲜度。 |
| mh_statusB | openmem 健康状态：总条数/钉住数/归档数/待审冲突/工具数/咨询次数。 |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 9 tools
The tools have distinct purposes, and the descriptions explicitly separate raw memory search (mh_search), synthesized answers (mh_ask), and pre-generated answers (mh_tool). Some risk remains because these three query-ish tools overlap in intent, but the guardrails in the descriptions are strong enough to guide selection.
All tool names use the consistent lowercase mh_ prefix and mostly follow a verb-based pattern like write, update, search, get, ask, and status. The noun-style exceptions mh_service, mh_tool, and mh_tools_list are minor deviations, and tools_list also reverses the expected order slightly.
Nine tools is well-scoped for a memory hub: core memory operations (write/get/search/update), specialized lookups (service/tool), and meta/status tools are all represented. Each tool has a clear role and none feel redundant.
The surface covers create, read, update, query, status, and specialized lookup workflows well. There is no explicit delete or conflict-resolution tool even though mh_status surfaces pending conflicts, but this may be intentional and can be worked around via archive/update and category changes.