@shaneholloman
基本情報
MCP server enabling persistent memory for Claude through a local knowledge graph - fork focused on local development
以下の設定を使って、このサーバーを MCP 対応クライアントに追加してください。
{
"mcpServers": {
"Aim-Memory-Bank": {
"command": "npx",
"args": [
"-y",
"mcp-knowledge-graph",
"--memory-path",
"/Users/yourusername/.aim"
]
}
}
}Store new memories. Use this to remember people, projects, concepts, or any information worth persisting. AIM (AI Memory) provides persistent memory for AI assistants. The 'aim_memory_' prefix groups all memory tools together. WHAT'S STORED: Memories have a name, type (person/project/concept/etc.), and observations (facts about them). DATABASES: Use the 'context' parameter to organize memories into separate graphs: - Leave blank: Uses the master database (default for general information) - Any name: Creates/uses a named database ('work', 'personal', 'health', 'research', etc.) - New databases are created automatically - no setup required - IMPORTANT: Use consistent, simple names - prefer 'work' over 'work-stuff' STORAGE LOCATIONS: Files are stored as JSONL (e.g., memory.jsonl, memory-work.jsonl): - Project-local: .aim directory in project root (auto-detected if exists) - Global: User's configured --memory-path directory - Use 'location' parameter to override: 'project' or 'global' RETURNS: Array of created entities. EXAMPLES: - Master database (default): aim_memory_store({entities: [{name: "John", entityType: "person", observations: ["Met at conference"]}]}) - Work database: aim_memory_store({context: "work", entities: [{name: "Q4_Project", entityType: "project", observations: ["Due December 2024"]}]}) - Master database in global location: aim_memory_store({location: "global", entities: [{name: "John", entityType: "person", observations: ["Met at conference"]}]}) - Work database in project location: aim_memory_store({context: "work", location: "project", entities: [{name: "Q4_Project", entityType: "project", observations: ["Due December 2024"]}]})
Link two memories together with a relationship. Use this to connect related information. RELATION STRUCTURE: Each link has 'from' (subject), 'relationType' (verb), and 'to' (object). - Use active voice verbs: "manages", "works_at", "knows", "attended", "created" - Read as: "from relationType to" (e.g., "Alice manages Q4_Project") - Avoid passive: use "manages" not "is_managed_by" IMPORTANT: Both 'from' and 'to' entities must already exist in the same database. RETURNS: Array of created relations (duplicates are ignored). DATABASE: Relations are created in the specified 'context' database, or master database if not specified. EXAMPLES: - aim_memory_link({relations: [{from: "John", to: "TechConf2024", relationType: "attended"}]}) - aim_memory_link({context: "work", relations: [{from: "Alice", to: "Q4_Project", relationType: "manages"}]}) - Multiple: aim_memory_link({relations: [{from: "John", to: "Alice", relationType: "knows"}, {from: "John", to: "Acme_Corp", relationType: "works_at"}]})
Add new facts to an existing memory. Use this to append information to something already stored. IMPORTANT: Memory must already exist - use aim_memory_store first. Throws error if not found. RETURNS: Array of {entityName, addedObservations} showing what was added (duplicates are ignored). DATABASE: Adds to entities in the specified 'context' database, or master database if not specified. EXAMPLES: - aim_memory_add_facts({observations: [{entityName: "John", contents: ["Lives in Seattle", "Works in tech"]}]}) - aim_memory_add_facts({context: "work", observations: [{entityName: "Q4_Project", contents: ["Behind schedule", "Need more resources"]}]})
Forget memories. Removes memories and their associated links. DATABASE SELECTION: Entities are deleted from the specified database's knowledge graph. LOCATION OVERRIDE: Use the 'location' parameter to force deletion from 'project' (.aim directory) or 'global' (configured directory). Leave blank for auto-detection. EXAMPLES: - Master database (default): aim_memory_forget({entityNames: ["OldProject"]}) - Work database: aim_memory_forget({context: "work", entityNames: ["CompletedTask", "CancelledMeeting"]}) - Master database in global location: aim_memory_forget({location: "global", entityNames: ["OldProject"]}) - Personal database in project location: aim_memory_forget({context: "personal", location: "project", entityNames: ["ExpiredReminder"]})
Remove specific facts from a memory. Keeps the memory but removes selected observations. DATABASE SELECTION: Observations are deleted from entities within the specified database's knowledge graph. LOCATION OVERRIDE: Use the 'location' parameter to force deletion from 'project' (.aim directory) or 'global' (configured directory). Leave blank for auto-detection. EXAMPLES: - Master database (default): aim_memory_remove_facts({deletions: [{entityName: "John", observations: ["Outdated info"]}]}) - Work database: aim_memory_remove_facts({context: "work", deletions: [{entityName: "Project", observations: ["Old deadline"]}]}) - Master database in global location: aim_memory_remove_facts({location: "global", deletions: [{entityName: "John", observations: ["Outdated info"]}]}) - Health database in project location: aim_memory_remove_facts({context: "health", location: "project", deletions: [{entityName: "Exercise", observations: ["Injured knee"]}]})
Remove links between memories. Keeps the memories but removes their connections. DATABASE SELECTION: Relations are deleted from the specified database's knowledge graph. LOCATION OVERRIDE: Use the 'location' parameter to force deletion from 'project' (.aim directory) or 'global' (configured directory). Leave blank for auto-detection. EXAMPLES: - Master database (default): aim_memory_unlink({relations: [{from: "John", to: "OldCompany", relationType: "worked_at"}]}) - Work database: aim_memory_unlink({context: "work", relations: [{from: "Alice", to: "CancelledProject", relationType: "manages"}]}) - Master database in global location: aim_memory_unlink({location: "global", relations: [{from: "John", to: "OldCompany", relationType: "worked_at"}]}) - Personal database in project location: aim_memory_unlink({context: "personal", location: "project", relations: [{from: "Me", to: "OldHobby", relationType: "enjoys"}]})
Read all memories in a database. Returns every stored memory and their links. FORMAT OPTIONS: - "json" (default): Structured JSON for programmatic use - "pretty": Human-readable text format DATABASE: Reads from the specified 'context' database, or master database if not specified. EXAMPLES: - aim_memory_read_all({}) - JSON format - aim_memory_read_all({format: "pretty"}) - Human-readable - aim_memory_read_all({context: "work", format: "pretty"}) - Work database, pretty
Search memories by keyword. Use this when you don't know the exact name of what you're looking for. WHAT IT SEARCHES: Matches query (case-insensitive) against: - Memory names (e.g., "John" matches "John_Smith") - Memory types (e.g., "person" matches all person memories) - Facts/observations (e.g., "Seattle" matches memories mentioning Seattle) VS aim_memory_get: Use aim_memory_search for fuzzy matching. Use aim_memory_get when you know exact names. FORMAT OPTIONS: - "json" (default): Structured JSON for programmatic use - "pretty": Human-readable text format EXAMPLES: - aim_memory_search({query: "John"}) - JSON format - aim_memory_search({query: "project", format: "pretty"}) - Human-readable - aim_memory_search({context: "work", query: "Shane", format: "pretty"})
Retrieve specific memories by exact name. Use this when you know exactly what you're looking for. VS aim_memory_search: Use aim_memory_get for exact name lookup. Use aim_memory_search for fuzzy matching or when you don't know exact names. RETURNS: Requested entities and relations between them. Non-existent names are silently ignored. FORMAT OPTIONS: - "json" (default): Structured JSON for programmatic use - "pretty": Human-readable text format EXAMPLES: - aim_memory_get({names: ["John", "TechConf2024"]}) - JSON format - aim_memory_get({names: ["Shane"], format: "pretty"}) - Human-readable - aim_memory_get({context: "work", names: ["Q4_Project"], format: "pretty"})
List all available memory databases and show current storage location. DATABASE TYPES: - "default": The master database (memory.jsonl) - used when no context is specified - Named databases: Created via context parameter (e.g., "work" -> memory-work.jsonl) RETURNS: {project_databases: [...], global_databases: [...], current_location: "..."} - project_databases: Databases in .aim directory (if project detected) - global_databases: Databases in global --memory-path directory - current_location: Where operations will default to Use this to discover what databases exist before querying them. EXAMPLES: - aim_memory_list_stores() - Shows all available databases and current storage location
Mcp Knowledge Graph is an MCP server that gives AI models persistent memory through a local knowledge graph. It stores and retrieves information across conversations using entities, relations, and observations, and works with Claude Code/Desktop and any MCP‑compatible AI platform.
Add the server to your claude_desktop_config.json or .claude.json using npx -y mcp-knowledge-graph --memory-path <path>. Optionally create a .aim directory in a project for project‑local storage. Once configured, the AI can use tools prefixed with aim_ (e.g., aim_memory_store, aim_memory_search) to manage memories.
.aim directories.aim directoryThe `.aim` directory is a project‑local folder that triggers automatic memory storage. The `_aim` marker is a safety line (`{"type":"_aim","source":"mcp-knowledge-graph"}`) placed at the start of every memory file to prevent accidental overwrites of unrelated JSONL files.
If a `.aim` directory exists in the current project, memory files are stored there. Otherwise, the configured global `--memory-path` directory is used. You can also force `project` or `global` location with an optional `location` parameter.
Pass a `context` parameter (e.g., `"work"`, `"personal"`) to memory tools. The system automatically creates new database files (e.g., `memory-work.jsonl`) alongside the master `memory.jsonl`. No manual setup is required.
The server refuses to write to the file and returns an error: “File does not contain required _aim safety marker”. Either add the marker manually or delete the file and let the system recreate it.
Node.js version 22 or later and an MCP‑compatible AI platform (e.g., Claude Desktop).
Groundwork: company memory for your AI tools. Docs public. Source closed. Public proof needs no account.
Personal context for every AI: search, read, and write back to your private Markdown library of articles, threads, PDFs, notes, and captured ChatGPT/Claude/Gemini/Grok conversations. Remote server, OAuth 2.1 paste-and-au
Islamic source data for assistants: Quran, hadith, tafsir, duas, prayer times and qibla. Answers come from real sources with citations instead of recalled text, so verses and hadith are not hallucinated. 36,000+ hadiths
Search published GPU deployment, rack integration and structured cabling resources; read canonical documents and Leviathan Systems company service scope. Anonymous, read-only Streamable HTTP MCP. No credentials required.
AI agents start every session with amnesia — you re-explain the project, repeat your preferences, and correct the same mistakes over and over. PLUR gives them a memory that persists. Your agent's corrections, preference
コメント