CogmemAi

Provides memory extraction, semantic search, and time-aware recall to Claude Code across sessions without local databases.

javascript

1

GitHub Stars

javascript

Language

4 months ago

First Indexed

2 months ago

Catalog Refreshed

Documentation & install

Readme and setup notes from the catalogue, plus a client-ready config you can copy for your MCP host.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "hifriendbot-cogmemai-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "cogmemai-mcp"
      ],
      "env": {
        "COGMEMAI_API_KEY": "cm_your_api_key_here",
        "COGMEMAI_API_URL": "https://api.example.com"
      }
    }
  }
}

CogmemAi MCP Server adds a persistent memory layer to Claude Code, enabling semantic memory, project-scoped recollection, and time-aware memory surfacing without local databases or complex infrastructure. It gives Claude Code a stable memory across sessions, so you can recall decisions, patterns, and preferences effortlessly.

How to use

You use the CogmemAi MCP Server by connecting Claude Code to the MCP endpoint via an MCP configuration. The server runs externally and handles memory extraction, semantic search, and memory surfacing. After you configure it, Claude Code will remember your architecture, preferences, decisions, and bugs across sessions without you having to prompt it each time.

How to install

Prerequisites: You need Node.js and npm installed on your machine.

# Install the MCP client globally
npm install -g cogmemai-mcp

Add CogmemAi MCP to Claude Code for per-project memory, or use a global configuration that applies to every project.

{
  "mcpServers": {
    "cogmemai": {
      "command": "cogmemai-mcp",
      "env": {
        "COGMEMAI_API_KEY": "cm_your_api_key_here"
      }
    }
  }
}

Or apply a global MCP configuration so every project automatically connects to CogmemAi without adding a project file.

claude mcp add-json cogmemai '{"command":"cogmemai-mcp","env":{"COGMEMAI_API_KEY":"cm_your_api_key_here"}}' --scope user

Additional setup notes

Restart Claude Code after configuring the MCP server. The memory layer will then be available across all sessions and projects.

Security and privacy notes

Memory facts are extracted and stored securely on CogmemAi’s side, with API keys hashed server-side and all traffic protected via HTTPS. No source code leaves your machine, and you can delete memories or disable the service at any time.

The system does not train models on your data.

Environment variables

The following environment variable is required to run CogmemAi MCP Server.

COGMEMAI_API_KEY=cm_your_api_key_here

How it works

During a session, CogmemAi extracts meaningful facts from your interactions, assigns semantic embeddings for memory items, and surfaces relevant memories at session start based on meaning, importance, and recency.

Works Everywhere

CogmemAi MCP Server works in any terminal where Claude Code runs, including PowerShell, bash/zsh, Windows Terminal, macOS Terminal, iTerm2, VS Code terminal, and any SSH session.

Tools and capabilities

CogmemAi provides tools to manage memories and interactions across sessions.

Notes and troubleshooting

If you encounter issues, ensure your API key is valid and that Claude Code can reach the CogmemAi MCP endpoint over HTTPS. If problems persist, regenerate your API key and reconfigure the MCP connection.

Available tools

save_memory

Store a fact explicitly such as an architecture decision or preference.

recall_memories

Search memories using natural language with semantic search.

extract_memories

Ai automatically extracts memories from a conversation.

get_project_context

Load memories relevant to the current project at session start.

list_memories

Browse memories with filters.

update_memory

Update a memory's content, importance, or scope.

delete_memory

Permanently delete a memory.

get_usage

Check your usage stats and plan information.