Kimi Code Memory MCP Server
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| MEMORY_STORE_ROOT | No | Override the storage root directory for memories |
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 | {} |
| prompts | {} |
| resources | {
"subscribe": false
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| rememberA | Write or overwrite a memory entry as a Markdown file with YAML frontmatter. |
| recallB | Read a memory entry by key and folder. |
| searchA | Case-insensitive keyword search across memory titles and contents. |
| list_tagsA | List all tags used in the current workspace. |
| listA | List memory entries in the workspace, sorted by most recently updated. |
| deleteA | Delete a memory entry by key and folder. |
| moveB | Move a memory entry to another folder, optionally renaming it. |
| load_more_contextA | Load older conversation rounds from the active wire.jsonl, summarized, before a given turn id. |
| search_contextA | Search conversation rounds across all workspace session wires by keywords and optional date range. Default detail: 'normal' keeps output within ~6000 chars. Use detail: 'compact' for a quick overview (no match text, no cluster members). Use detail: 'full' when you need full match text and all cluster members. |
| list_search_viewsA | List saved search views. Each view records the clusters discovered by a previous search_context call. Use these views as candidate sets before creating or extending a theme. |
| delete_search_viewA | Delete a saved search view. Set deleteRefinedTurns to true to also remove the refined turns referenced by this view (useful for purging low-quality refined data). |
| load_turn_contextA | Load the full detailed content of specific conversation turns by sessionId and turnId. |
| tag_themeA | 仔细分析 turn 内容与 theme 确定相关后，将 turn 挂载到 theme。禁止仅凭关键词匹配挂载；必须确认内容 genuinely belongs to the theme 才可关联。如果 theme 不存在会自动创建。 |
| trace_themeA | Trace the evolution of a theme across sessions and memories. Returns associated turns and memories sorted by time. |
| list_themesA | List all theme identifiers stored in the current workspace. |
| delete_themeA | Delete a theme association file. This does not delete the refined turns or memories referenced by the theme. |
| refine_session_turnsA | Read a session wire.jsonl and generate Refined Turn Summaries. Output is written to refined/.jsonl. |
| get_current_workspaceA | Return the current cwd, workspace id and store path. |
| open_memory_dashboardA | Opens the memory dashboard in the default browser. Starts the dashboard server if it is not already running. |
| organize_memoriesB | Two-stage workspace memory organizer. Empty call returns existing essence + pending memory files + rules. Call with content to store the organized essence.md. |
| sync_workspace_indexA | Reconciles index.json with the filesystem. Empty call scans and reports mismatches. Call with folderComments to set folder descriptions. |
| bootstrap_workspaceB | Session bootstrap: loads workspace context, essence, notes refs, and a memory index tree with recent changes marked [new]. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
| memory-decision-check | Before modifying a file, check historical decisions related to it. |
| memory-theme-trace | Trace the evolution of a theme across sessions. |
| memory-session-summary | Summarize the current session and suggest themes to tag. |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 22 tools
Tools cluster into clear groups (memory CRUD, context loading, theme management, workspace maintenance) and each has a distinct action/target. Some pairs like search vs search_context or organize_memories vs bootstrap_workspace could require careful reading, but descriptions resolve the boundaries.
Almost all names follow a lowercase verb or verb_noun pattern (remember, list_tags, trace_theme, sync_workspace_index). A few generic single-verb names like list, delete, and move are less descriptive but still stylistically consistent.
At 22 tools, this sits in the heavy range, and the server spans several subdomains (memory, context, themes, workspace admin). Each tool has a plausible purpose, but the set feels broader than a tightly scoped memory server.
Memory entries have full lifecycle coverage (create/read/search/list/update via overwrite/delete/move), and themes/context are well supported. Minor gaps remain around removing a single turn from a theme or enumerating sessions, but agents can generally accomplish core workflows.