openmemkit
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": false
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| memory_bootstrapA | Call once at session start. Returns MEMORY.md rules, all domain indexes, environment paths, and semantic-search status. / 会话开始时调用一次。返回 MEMORY.md 规则、全部域索引、环境路径与语义检索状态。 |
| memory_domainsA | List memory domains with their .md file counts. / 列出记忆库各域及其 .md 文件数量。 |
| memory_searchA | Search memory chunks. mode: keyword (FTS5/LIKE, default for short queries), hybrid (keyword + vector RRF fusion, default), vector (pure cosine). Filter by domain/date_range. / 检索记忆正文。mode=keyword(关键字)、hybrid(关键字+向量融合,默认)、vector(纯向量)；支持域/日期过滤。 |
| memory_getA | Read a single .md file by path (relative to memory root). Top-level MEMORY.md by bare name; domain files like web/2026-08-20.md. / 按 path 精读单个 .md 文件。顶层 MEMORY.md 用裸名；域文件如 web/2026-08-20.md。 |
| memory_listB | List indexed files with domain, chunk count, mtime. Supports domain filter and limit. / 列出已索引文件，含域、chunk 数、修改时间；支持域过滤与限量。 |
| memory_statsA | Return index statistics: chunk count, per-domain distribution, write log status, semantic coverage, db size. / 返回索引统计：chunk 数、域分布、write_log 状态、语义覆盖、数据库体积。 |
| memory_updateA | Replace the content of an existing .md file. The previous version is archived under .archive/ and the change is recorded in the audit log. Refuses MEMORY.md. / 替换已有 .md 文件内容；旧版本归档到 .archive/，变更记入审计日志。禁止更新 MEMORY.md。 |
| memory_deleteA | Delete a .md file by moving it to .trash/ (recoverable) and recording a tombstone in the audit log. Refuses MEMORY.md. / 删除 .md 文件：移入 .trash/（可恢复）并写 tombstone 审计记录。禁止删除 MEMORY.md。 |
| memory_historyA | Show the audited change history for a path (or the whole write_log). / 查看某路径（或整个 write_log）的审计变更历史。 |
| memory_writeA | Append to the audited write_log (does NOT modify .md directly). network_fetch entries auto-apply to web/ on flush; other kinds are distributed per config. / 写入审计式 write_log（不直接改 .md）。network_fetch 类条目在 flush 时自动落 web/，其余按配置分发。 |
| memory_flushA | Trigger immediate distribution of pending write_log entries to .md files, then reindex. / 触发立即分发 pending 条目到 .md 文件并重建索引。 |
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 11 tools
memory_domains, memory_list, and memory_stats all return overlapping inventory/index information, which could cause misselection. memory_update and memory_write also both suggest writing, though they target different backends; the descriptions help, but boundaries are not always obvious.
All tools share the memory_ prefix and most follow a verb-oriented pattern like search, get, list, update, delete, write, and flush. A few noun-style names such as domains, stats, and history deviate slightly, but the overall convention remains predictable.
11 tools is well within the ideal range for a memory-management server. Each major operation—bootstrap, read, search, list, stats, update, delete, history, write, and flush—has a dedicated tool without excessive fragmentation.
The set covers session bootstrap, reading/searching, metadata inspection, update/delete, audit history, and the write_log flush pipeline. Minor gaps exist: delete is described as recoverable but no restore tool is exposed, and there is no direct create-file operation apart from the write_log/flush path.