Build Persistent Memory for AI Agents with TypeScript, SQLite & MCP (2026 Guide)

3. Step-by-Step Implementation

Install dependencies

npm install @modelcontextprotocol/sdk better-sqlite3 zod
npm install -D @types/better-sqlite3 typescript

Database schema (src/memory/db.ts)

import Database from 'better-sqlite3'
import { drizzle } from 'drizzle-orm/better-sqlite3'

const sqlite = new Database(process.env.MEMORY_DB_PATH ?? './agent-memory.db')
sqlite.pragma(`journal_mode = WAL`)
sqlite.pragma(`foreign_keys = ON`)

sqlite.exec(`
  CREATE TABLE IF NOT EXISTS memories (
    id TEXT PRIMARY KEY,
    title TEXT NOT NULL,
    content TEXT NOT NULL,
    type TEXT DEFAULT 'observation',
    project TEXT,
    created_at TEXT DEFAULT (datetime('now')),
    updated_at TEXT DEFAULT (datetime('now'))
  );
  CREATE VIRTUAL TABLE IF NOT EXISTS memories_fts USING fts5(
    title, content, content='memories', content_rowid='rowid'
  );
`)

export const db = sqlite

Core memory tools (MCP)

import { McpServer } from '@modelcontextprotocol/sdk/server/mcp.js'
import { StdioServerTransport } from '@modelcontextprotocol/sdk/server/stdio.js'
import { z } from 'zod'
import { db } from './db.js'
import { randomUUID } from 'crypto'

const server = new McpServer({ name: 'agent-memory', version: '1.0.0' })

server.tool(
  'memory_create',
  'Store a new observation or decision for later retrieval',
  {
    title: z.string().describe('Short title'),
    content: z.string().describe('Full observation or decision text'),
    type: z.string().optional().default('observation'),
    project: z.string().optional()
  },
  async ({ title, content, type, project }) => {
    const id = randomUUID()
    db.prepare(
      `INSERT INTO memories (id, title, content, type, project) VALUES (?, ?, ?, ?, ?)`
    ).run(id, title, content, type, project ?? null)
    return { content: [{ type: 'text', text: `Stored memory ${id}` }] }
  }
)

server.tool(
  'memory_search',
  'Search past memories by keyword. Returns compact index with IDs.',
  {
    query: z.string().describe('Search terms'),
    limit: z.number().optional().default(5)
  },
  async ({ query, limit }) => {
    const rows = db.prepare(
      `SELECT m.id, m.title, snippet(memories_fts, 1, '', '', '...', 20) as snippet
       FROM memories_fts f JOIN memories m ON f.rowid = m.rowid
       WHERE memories_fts MATCH ? ORDER BY rank LIMIT ?`
    ).all(query, limit)
    return { content: [{ type: 'text', text: JSON.stringify(rows, null, 2) }] }
  }
)

const transport = new StdioServerTransport()
await server.connect(transport)

Run the server with npx tsx src/memory/server.ts and add it to your agent's MCP config.