Pg-Mnemosyne MCP

A Model Context Protocol (MCP) server that provides AI assistants with a robust "super memory", task tracker, and dynamic PostgreSQL database management capabilities.

⚡ Quick Start

1. Install the package globally:

   • Windows:
     Terminal

     Copy

     pip install pg-mnemosyne-mcp
     

   • macOS / Linux:
     bash

     Copy

     pipx install pg-mnemosyne-mcp
     

     (If you prefer standard pip or don't have pipx, run: pip install pg-mnemosyne-mcp --break-system-packages)

2. Auto-configure all your AI agents (Claude, Gemini, Qwen, Cursor, etc.) at once:

   bash

   Copy

   pg-mnemosyne init --dsn "postgresql://user:password@localhost:5432/postgres"
   

   (Be sure to replace user and password with your actual PostgreSQL username and database password!)

3. Restart your AI agents. You're done!

Features

• High-Performance: Uses cached connection pooling (asyncpg.create_pool) for instant sub-millisecond database queries.
• Dynamic Projects: The AI can create new databases for different projects on the fly.
• Dynamic Schema: The AI can modify table schemas dynamically to adapt to changing memory needs.
• Standard Memory Tracker: Built-in support for tracking, updating, and deleting memory items with tags.
• Advanced Task Management: Dedicated tasks schema with fields for status transitions, priority, and deadlines.
• Multi-Agent Coordination: Centralized session tracking preventing duplicate coding and file-editing conflicts.
• Raw SQL Execution: Gives AI ultimate flexibility for complex queries and DDL operations.

Setup

Users will need to provide their PostgreSQL credentials using the PG_BASE_DSN environment variable. This is a standard connection string: postgresql://<USERNAME>:<PASSWORD>@<HOST>:<PORT>/<DEFAULT_DB>

Where to configure this (Client Setup)

The exact location depends on which AI client you are using. You need to add the server configuration to your client's MCP settings file.

For Claude Desktop:

• Mac: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json

For Cursor:

• Go to Settings > Features > MCP and add a new MCP server, or edit your project's .cursor/mcp.json.

For Roo Code / Cline (VS Code):

• Edit the MCP settings file located at ~/Library/Application Support/Code/User/globalStorage/saoudrizwan.claude-dev/settings/cline_mcp_settings.json (Mac) or the equivalent Windows path.

For Gemini CLI & Qwen CLI:

• Open your global configuration file (usually located at ~/.gemini/settings.json or ~/.qwen/settings.json).
• Alternatively, use the CLI:
  bash

  Copy

  gemini mcp add pg-mnemosyne "/path/to/pg-mnemosyne" -e PG_BASE_DSN="postgresql://user:pass@localhost:5432/postgres" -s user
  # OR
  qwen mcp add pg-mnemosyne "/path/to/pg-mnemosyne" -e PG_BASE_DSN="postgresql://user:pass@localhost:5432/postgres" -s user
  

For Claude Code CLI:

• The easiest way is to add it via the CLI:
  Terminal

  Copy

  claude mcp add pg-mnemosyne "/path/to/pg-mnemosyne" -e PG_BASE_DSN="postgresql://user:pass@localhost:5432/postgres" -s user
  

• Manually, it lives in your global config at ~/.claude.json.

For Codex CLI:

• The easiest way is to add it via the CLI:
  bash

  Copy

  codex mcp add pg-mnemosyne --env PG_BASE_DSN="postgresql://user:pass@localhost:5432/postgres" -- pg-mnemosyne
  

• Manually, it lives in your global config at ~/.codex/config.toml (TOML format).

For Windsurf IDE:

• Edit your global config at ~/.codeium/windsurf/mcp_config.json.
• Alternatively, click the Hammer (MCP) icon in the Cascade panel and select Configure.

For Antigravity CLI (agy): Antigravity uses a plugin-based system. To add the server:

1. Create a plugin directory: mkdir -p ~/.gemini/config/plugins/pg-mnemosyne
2. Create ~/.gemini/config/plugins/pg-mnemosyne/mcp_config.json with the Standard Template below.
3. Add an entry to your ~/.gemini/config/import_manifest.json under the "imports" array:
   config.json

   Copy

   {
     "name": "pg-mnemosyne",
     "source": "manual",
     "components": ["mcpServers"]
   }
   

Configuration Template (Claude Desktop, Cursor, Roo Code, Gemini CLI, Claude Code, Antigravity, Windsurf):

{
  "mcpServers": {
    "pg-mnemosyne": {
      "command": "pg-mnemosyne",
      "env": {
        "PG_BASE_DSN": "postgresql://postgres:my_password@localhost:5432/postgres"
      }
    }
  }
}

For OpenCode:

• Edit your OpenCode configuration file located at ~/.config/opencode/opencode.jsonc.

Configuration Template (OpenCode):

{
  "mcp": {
    "pg-mnemosyne": {
      "type": "local",
      "command": ["pg-mnemosyne"],
      "environment": {
        "PG_BASE_DSN": "postgresql://postgres:my_password@localhost:5432/postgres"
      }
    }
  }
}

Running the Server (Standalone)

pg-mnemosyne run

This starts the MCP server using standard input/output.

CLI Usage

The pg-mnemosyne command also acts as a standalone CLI for managing your data and configuring agents.

Auto-Initialization

You can automatically configure all supported AI agents (Claude, Gemini, Qwen, Cursor, etc.) with a single command:

pg-mnemosyne init --dsn "postgresql://user:pass@localhost:5432/postgres"

Manual Record Management

You can add and list records directly from your terminal:

# Add a record
pg-mnemosyne add my_project_db todo "Finish the documentation"

# List records
pg-mnemosyne list my_project_db --type todo

🤝 Multi-Agent Coordination & Advanced Tasks

Pg-Mnemosyne includes specialized schemas to help complex multi-agent setups (e.g. Gemini CLI, Codex CLI, Roo Code, Claude Desktop) coordinate on the same project:

📋 Professional Tasks Schema

Spin up a dedicated tasks table with fields for statuses (backlog, todo, in_progress, blocked, done), priority levels (low, medium, high, critical), tags, and deadlines:

pg-mnemosyne init-todo my_project_db

🛰️ Agent Coordination Hub

Avoid merge conflicts, double-coding, and redundant compiler troubleshooting by initializing the shared agent_sessions coordination table:

pg-mnemosyne init-coordination my_project_db

When active, agents use the update_agent_session and get_active_sessions MCP tools to register their current editing files and active subtasks, creating a real-time bulletin board for mutual visibility!

Available MCP Tools

• create_project_db(db_name: str): Creates a new isolated PostgreSQL database.
• init_schema(db_name: str): Initializes the base records table.
• init_todo_schema(db_name: str): Initializes a professional tasks table.
• init_coordination_schema(db_name: str): Initializes the multi-agent agent_sessions table.
• add_column(db_name: str, table: str, column_name: str, data_type: str): Dynamically adds a column to any table.
• add_record(db_name: str, type: str, content: str, tags: list[str]): Adds a memory/todo record.
• get_records(db_name: str, type: str = None, limit: int = 50): Retrieves recent records.
• update_record(db_name: str, record_id: int, content: str = None, tags: list[str] = None, status: str = None): Partially updates a record.
• delete_record(db_name: str, record_id: int): Deletes a record by ID.
• update_agent_session(db_name: str, agent_name: str, active_task: str, active_file: str = None, status: str = "active"): Registers/updates active agent state.
• get_active_sessions(db_name: str): Lists active agent coordination sessions.
• run_sql(db_name: str, query: str): Runs arbitrary SQL (SELECT, INSERT, DDL, etc.).