Practical overview
When it fits
Use it when an MCP-compatible assistant needs persistent, locally stored information beyond the current conversation, especially when project-local and global memory locations or topic-specific databases are useful.
Who it is for
- People configuring MCP-compatible assistants that need memory across conversations.
- Developers who want to manage local assistant memory as entities and relationships.
- Users who want separate project, work or personal memory databases.
Documented uses
- Save facts about people, projects or concepts for retrieval in later conversations.
- Connect related memories and search them by keyword or exact name.
- Separate memory by project or named context, and choose where the files are stored.
- Review available memory databases and remove outdated memories or facts.
Requirements & limits
The README specifies Node.js 22 or newer and an MCP-compatible AI platform. It documents Claude Desktop and Claude Code configuration examples and identifies the project as MIT-licensed.
- A project-root .aim directory takes precedence over the configured global path.
- The server stores memory in JSONL files and requires its _aim marker before writing.
- Named contexts use separate database files; the default master database is used when no context is specified.
- A synced folder can be selected as the storage path for cross-machine access.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Configure a global memory directory
Add the MCP server to an assistant configuration with an explicitly chosen global storage path.
Before you start
- Node.js 22 or newer
- An MCP-compatible AI platform
- Access to the platform’s MCP configuration file
Steps
Choose a directory for memory files, such as a local .aim directory.
Add an mcpServers entry for mcp-knowledge-graph to the platform configuration.
Set the command to npx and include the package name and --memory-path followed by the chosen directory.
Expected outcome
The configured server uses the chosen directory for its default memory database and any named context databases.
Guide limitations
- Configuration details vary by host; the README gives examples for Claude Desktop and Claude Code.
- A project containing a .aim directory takes priority for project-local storage.
Set up project-local memory
Use a project-specific master memory file by creating the directory the server detects.
Before you start
- Node.js 22 or newer
- An MCP-compatible AI platform configured to use the server
- A project directory
Steps
From the project root, create a directory named exactly .aim.
Run the configured memory tools from that project context.
Expected outcome
The server uses .aim/memory.jsonl as the project-local master database.
Guide limitations
- The directory must be named exactly .aim for project detection.
- This guide covers storage setup; it does not describe configuring every MCP host.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 12 October 2026
- Original website title
- shaneholloman/mcp-knowledge-graph · Official homepage title · retained in its source language