mem-plain
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| MEMORY_DIR | No | Directory where the plain-text memory files (FACT.md, JOURNAL.jsonl, SOUL.md, USER.md) are stored. Change it to move memory to another location. | E:\memory |
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": false
} |
| prompts | {
"listChanged": false
} |
| resources | {
"subscribe": false,
"listChanged": false
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| memory_readA | 读取记忆文件。filename 默认 FACT.md；可传 SOUL.md / USER.md / JOURNAL.jsonl / FACT.archive.md（归档）。三种取法（按优先级）：tail>0 只返回最后 tail 行；offset>0 或 limit>0 按行号读一段（offset 从 1 起，limit 最多几行；配合 memory_search 给的行号读归档最省 token）；都不给就整份返回。 |
| memory_updateA | 覆盖写入 FACT.md（长期记忆）。写入前会自动备份，并有缩水保护：若新内容不足 1KB 且旧内容的三倍以上，则拒绝（防写坏）。 |
| memory_appendA | 往 JOURNAL.jsonl 追加一条事件记录。tags 可选，是标签列表。默认先做「巩固」：如果今天已有条目与新内容相似度 ≥90%（或新内容被旧条目包含），就不再新增，而是把新信息并进那条（保留原时间戳、tags 取并集）。merge=False 可强制新增。 |
| memory_searchA | 在记忆里检索。默认同时搜 JOURNAL 事件和归档文件（FACT.archive.md）——归档是从 FACT 搬走的旧段落，不占固定注入，靠这里找回。query 为空时只按 tag / 日期过滤 JOURNAL（= 列出某段时间的事）。tag 可写多个（逗号分隔），命中任一即可；after / before 用 YYYY-MM-DD（本地日期，含当天）。返回最近 limit 条事件（带时间戳与 tags）和命中的归档行。大小写不敏感的普通子串匹配，不是语义检索。 |
| memory_statsA | 记忆体检（只读）：返回各记忆文件的体积、行数/条数、最后修改时间与距今天数，JOURNAL 的首末时间与今日新增条数，以及备份文件占用。适合在整理记忆前、或回答「你记不记得某件事」之前先看整体状态——它只量尺寸，不判断内容重要性、不动任何文件。 |
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 5 tools
The five tools mostly map to distinct operations: read a file, append a journal event, search across memory, inspect stats, and overwrite FACT.md. The only mild overlap is between memory_read (which can also read FACT.archive.md) and memory_search (which surfaces archived lines), but the descriptions make the distinction clear.
All tool names follow the exact same `memory_<verb>` snake_case pattern. The verbs (read, append, search, stats, update) are predictable and readable.
Five tools is well-scoped for a personal memory server: one per core operation with nothing redundant. Each tool earns its place.
Read, append, search, stats, and overwrite are covered, but there is no way to delete or prune JOURNAL entries and no exposed mechanism to archive FACT.md into FACT.archive.md despite the archive file being a first-class concept. Agents can work around deletions by editing files externally, but the lifecycle surface has notable gaps.