Provides memory management tools for OpenAI Codex, enabling AI agents to store, search, retrieve, update, and delete memories via the mem9 memory service.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@mem9 MCP Server remember that my favorite color is blue and I like hiking"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Universal MCP Server for mem9 memory service. One server, all platforms.
# Requires a running mnemo-server and API key
git clone https://github.com/you06/mem9-mcp.git
cd mem9-mcp
npm installVariable | Required | Default | Description |
| Yes | - | API key for authentication (also serves as tenant ID) |
| No |
| mnemo-server base URL |
| No | auto-detected | Agent identifier |
| No |
| Log level: debug, info, warn, error |
| No |
| HTTP request timeout (ms) |
| No |
| Default search result limit |
Agent ID is auto-detected from platform environment variables (CURSOR_WORKSPACE, CLAUDE_CODE_VERSION, CODEX_CLI_VERSION).
Related MCP server: Simple Memory MCP
Tool | Description |
| Store a new memory (processed via fact extraction) |
| Search memories by semantic similarity and keywords |
| Retrieve a specific memory by ID |
| Update an existing memory (direct field update) |
| Delete a memory |
npm install
npm run build
npm testBefore publishing, you can test the MCP server locally by pointing to the built entry file:
Claude Code:
claude mcp add mem9 -e MEM9_API_KEY=your-key -e MEM9_API_URL=http://localhost:8080 -- node /path/to/mem9-mcp/dist/index.jsOr in .mcp.json:
{
"mcpServers": {
"mem9": {
"command": "node",
"args": ["/path/to/mem9-mcp/dist/index.js"],
"env": {
"MEM9_API_URL": "http://localhost:8080",
"MEM9_API_KEY": "your-key"
}
}
}
}Codex:
codex mcp add mem9 -- node /path/to/mem9-mcp/dist/index.jsOr in .codex/config.toml:
[mcp_servers.mem9]
command = "node"
args = ["/path/to/mem9-mcp/dist/index.js"]
[mcp_servers.mem9.env]
MEM9_API_URL = "http://localhost:8080"
MEM9_API_KEY = "your-key"After code changes, run npm run build and restart the MCP client to pick up the new build.
Settings > Features > MCP Servers:
{
"mcpServers": {
"mem9": {
"command": "npx",
"args": ["-y", "@mem9/mcp-server"],
"env": {
"MEM9_API_URL": "http://localhost:8080",
"MEM9_API_KEY": "your-key"
}
}
}
}claude mcp add mem9 -- npx -y @mem9/mcp-serverOr in .mcp.json:
{
"mcpServers": {
"mem9": {
"command": "npx",
"args": ["-y", "@mem9/mcp-server"],
"env": {
"MEM9_API_URL": "http://localhost:8080",
"MEM9_API_KEY": "your-key"
}
}
}
}codex mcp add mem9 -- npx -y @mem9/mcp-serverOr in .codex/config.toml:
[mcp_servers.mem9]
command = "npx"
args = ["-y", "@mem9/mcp-server"]
[mcp_servers.mem9.env]
MEM9_API_URL = "http://localhost:8080"
MEM9_API_KEY = "your-key"Apache-2.0
Delete a memory that is no longer relevant or accurate. Use this to clean up outdated information. This action is irreversible.
Inputs: id (required).
Retrieve a specific memory by its ID. Use this when you have a memory ID from a previous search and need the full content.
Inputs: id (required).
Search your memory for relevant information. Use this at the start of a task to recall relevant context, when you need to check if something was previously discussed, or when the user references past work. Supports both semantic (meaning-based) and keyword search.
Inputs: query (required), tags, limit.
Store a new memory. The content will be processed asynchronously by the memory service (fact extraction and reconciliation) — it is not stored verbatim and may take a few seconds to become searchable. Use this when you learn something worth remembering: user preferences, project conventions, important decisions, recurring patterns, or any context that would be useful in future sessions. Do NOT store trivial or transient information like current file paths, temporary debug notes, or information that only matters in the current session. Note: the response does not include the stored memory's ID. If you need to update or delete a memory you just stored, use memory_search to find it first.
Inputs: content (required), tags, metadata, session_id.
Update an existing memory. Use this when information has changed, needs correction, or should be enriched with additional context. Provide the memory ID and at least one field to update. This is a direct field update — content is NOT re-processed through reconciliation.
Inputs: id (required), tags, content, metadata.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 5 tools
Each tool targets a distinct operation: create (store), read (get/search), update, and delete. There is no overlap between retrieving by ID, searching, storing, updating, or deleting.
All tool names follow the same verb_noun pattern: memory_get, memory_delete, memory_store, memory_search, memory_update. Consistent and predictable.
Five tools is well-scoped for a memory server. It covers all core operations without unnecessary bloat or missing essentials.
The tool set provides full CRUD coverage (store, get, update, delete) plus search, which is the complete lifecycle for a memory service. No obvious gaps.
Persistent personal memory for AI assistants — save, search, and recall across every MCP client.
Persistent, verbatim, searchable memory for AI assistants — one memory across every MCP client.
Securely search, create, and organize your Mem notes and collections from AI assistants.
Persistent memory for AI assistants: store, search, and connect knowledge across conversations.
