global-memory
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| GLOBAL_MEMORY_AI | No | Pin the AI provider used for distillation and consolidation: 'claude', 'codex', or 'gemini' (auto-detected by default). | |
| GLOBAL_MEMORY_DIR | No | Data directory for the preset store, skills, and backups (default ~/.global-memory). | ~/.global-memory |
| GLOBAL_MEMORY_HOME | No | Fake HOME from which all client paths are derived, for testing purposes. | |
| GLOBAL_MEMORY_CONSENT | No | Consent behavior: 'auto' (default) uses dialogs on local macOS sessions, 'dialog' forces dialogs, 'conversation' falls back to in-conversation confirmation. | auto |
| GLOBAL_MEMORY_RAW_DIR | No | Output directory for 'extract-raw' command (default <repo>/raw-memories). | |
| GLOBAL_MEMORY_BATCH_CAP | No | Maximum number of presets per checklist dialog (default 8). | 8 |
| GLOBAL_MEMORY_NO_AUTO_SYNC | No | Set to '1' to disable self-healing on server start (e.g., restore missing skill links or hooks). | 0 |
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 |
|---|---|
| recall_presetsA | 开始任务时调用，召回适用于当前上下文的预设并完成确认（服务器可能向用户弹系统确认框，等待返回即可）。返回终态：applied=已应用（照单执行）、skipped=不要应用、pending_confirmation=请在对话中逐条向用户确认（结果用 record_usage 回报）、backlog=待补确认（方便时向用户提及）、suggestions=单条转达的系统建议。返回的 user_notice 必须原样转述给用户。Recall and resolve applicable presets; relay user_notice verbatim. |
| save_requirementA | 把用户确认过的可复用要求保存为预设。调用前必须已向用户展示提炼表述+建议分类并获明确同意（弹框通道下服务器会再次弹框担保，超时不保存）。单次指令不入库。返回 potential_conflicts=同作用域/分类的既有预设——检查语义矛盾，矛盾则提议更新旧预设而非新增。user_notice 原样转述。Save a user-confirmed durable requirement. |
| record_usageA | 回报对话中完成的预设确认结果（pending_confirmation 的逐条答复、backlog 补确认、纠正流程的推翻）。confirmed=用户确认沿用；rejected=用户拒绝；overridden=应用后被用户当场推翻。Report per-preset consent outcomes resolved in conversation. |
| list_memoriesB | 列出或搜索已保存的预设。所有过滤条件可选。List/search saved presets. |
| update_memoryA | 修改预设：内容、作用域、分类、策略、启用/停用/归档。需用户已确认该修改。未传字段不变；project/task_type/category 传 "" 清除。内容实质改写会重置信任（新表述=未验证的新承诺）。user_notice 原样转述。Update a preset (user-approved). |
| delete_memoryA | 永久删除一条预设。必须先向用户复述内容并获明确确认；临时不用应改为 status=disabled。Permanently delete (confirm with user first; prefer disabling). |
| memory_overviewA | 统计 + 体检指标（疑似重复/陈旧候选/毕业候选簇/待补确认/上次整理时间）。用户说"看看我的记忆"时先调这个。Overview with health checks. |
| start_consolidationA | 仅在用户主动要求整理记忆时调用。返回全库与体检指标。你的任务：产出整理提案清单（合并语义重复、解决矛盾、提炼表述、陈旧停用、聚类毕业成 skill），以清单呈现、由用户驱动决定，再用 apply_consolidation 落地。Start a user-initiated consolidation pass. |
| apply_consolidationA | 原子地执行用户已确认的整理操作清单。仅限用户主导的管理场景。op: update（带 fields）/ delete / disable / archive（可带 note 注明去向）/ create（带 fields）。Atomically apply user-approved consolidation operations. |
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
Most tools have clearly distinct roles: recall/record_usage, list/overview, and start_consolidation/apply_consolidation are well separated. The only potential confusion is between update_memory and apply_consolidation (both can modify presets) and between delete_memory and apply_consolidation's delete, but the usage context (direct user action vs. consolidation workflow) disambiguates them.
Eight of nine tools follow a consistent verb_noun pattern (record_usage, list_memories, update_memory, start_consolidation, apply_consolidation, recall_presets, save_requirement, delete_memory). memory_overview breaks the pattern by being noun_noun, but it is still readable and clearly the odd one out.
Nine tools is a well-scoped size for a memory/preset management server. Each tool covers a distinct operation or workflow stage—CRUD, recall, usage reporting, and consolidation—without redundancy or bloat.
The tool surface covers the full lifecycle: save_requirement creates presets, list_memories/memory_overview/recall_presets read them, update_memory plus apply_consolidation handle updates/disable/archive, and delete_memory covers deletion. The two-phase consolidation workflow and record_usage for outcome tracking round out the domain with no obvious gaps.