Amnesic

@SurajKGoyal

About Amnesic

Persistent semantic memory for SQL databases (Postgres, MySQL, MSSQL, SQLite). Annotate once, remember forever.

#database #sql #postgres #mysql #mssql #sqlite #memory #knowledge #annotations #claude #cursor

Config

Add this server to your MCP-compatible client using the configuration below.

{
  "mcpServers": {
    "amnesic": {
      "command": "uvx",
      "args": [
        "--from",
        "amnesic[all]",
        "amnesic"
      ]
    }
  }
}

Tools

List configured connections (no secrets)

All known tables with descriptions

BM25 search over table/column knowledge

Column schema merged with saved annotations

Execute a read-only SELECT

Persist semantic annotations (enum meanings, FKs)

Copy annotations between connections (staging → prod)

Discover all FK relationships from the live DB

Navigate the FK graph for JOIN planning

Overview

What is Amnesic?

Amnesic is an MCP server that provides persistent semantic memory for SQL databases, allowing AI sessions to retain knowledge about table schemas, enum meanings, and foreign key relationships across sessions. It is intended for developers and data professionals who want to avoid re-explaining database context when interacting with AI tools.

How to use Amnesic?

Install with pipx install amnesic and run amnesic init to set up connections. The server exposes nine MCP tools for listing connections, querying schemas, executing read‑only SELECT queries, annotating database semantics, and managing knowledge across connections.

Key features of Amnesic

Persistent SQLite knowledge store per database
Read‑only by design with two independent safety layers
Supports PostgreSQL, MySQL/MariaDB, SQL Server, and SQLite
Credentials never appear in tool responses
Nine MCP tools for schema, query, and annotation management

Use cases of Amnesic

Annotating enum values once so every future AI session sees the labels
Sharing database context across multiple AI tools or sessions
Copying semantic annotations from staging to production synchronously
Discovering foreign key relationships for automated JOIN planning

FAQ from Amnesic

How does Amnesic ensure queries are safe for production?

It uses two independent layers: static SQL analysis rejects any write/DDL statement before connecting, and every query runs inside a transaction that is immediately rolled back.

What databases does Amnesic support?

PostgreSQL, MySQL/MariaDB, Microsoft SQL Server, and SQLite.

Do credentials ever appear in tool responses?

No, credentials are never exposed in tool responses.

Can I copy annotations between environments?

Yes, the db_sync_knowledge tool copies annotations between connections (e.g., from staging to production).

Where is the semantic knowledge stored?

In a local SQLite knowledge store, one per database, that persists across sessions.

Frequently asked questions

How does Amnesic ensure queries are safe for production?

It uses two independent layers: static SQL analysis rejects any write/DDL statement before connecting, and every query runs inside a transaction that is immediately rolled back.