THE DATABASE
SYSTEM FOR AI
APPLICATIONS
EVERY
AGENT
EVERY
TOOL CALL
EVERY
CONTEXT
INFINITELY
SCALABLE
Agents create 100x more databases than humans: instantly provision with just a unique ID, no setup, no configuration.
MINIMIZED
COST
AI systems are expensive enough, only pay for queries and storage of databases made by agents at scale.
DATA
PROVENANCE
Keep databases isolated for easy download, reuse as context, and deletion for compliance.

USAGE SAMPLES

An agent decides it needs a database to complete its task

It creates one simply by generating a unique ID

The agent then creates the schema for the tables it needs

The database is stored as a file so no need to manage compute.

Each database is isolated and can be reused, downloaded, deleted, etc.

Upload or sync any number of exiting databases

AgentDB is also an MCP server

Agents can then use databases as context

CODE SAMPLES 

import { DatabaseService } from "@agentdb/sdk";
import fs from "fs";
import "dotenv/config";

const agentdb = new DatabaseService("https://api.agentdb.dev", process.env.AGENTDB_API_KEY, true);

// Generate a new UUID for the database token
const newDatabaseToken = crypto.randomUUID();
console.log(`Generated new database token: ${newDatabaseToken}`);

const connection = agentdb.connect(newDatabaseToken, "users_db", "sqlite");

const main = async () => {
  try {
    console.log("Creating users table...");

    // Create the users table
    await connection.execute({
      sql: `CREATE TABLE IF NOT EXISTS users (
        id INTEGER PRIMARY KEY AUTOINCREMENT,
        name TEXT NOT NULL,
        email TEXT UNIQUE NOT NULL,
        age INTEGER,
        created_at DATETIME DEFAULT CURRENT_TIMESTAMP
      )`
    });

    console.log("Populating users table with sample data...");

    // Insert sample data
    const sampleUsers = [
      { name: "Alice Johnson", email: "alice@example.com", age: 28 },
      { name: "Bob Smith", email: "bob@example.com", age: 34 },
      { name: "Carol Davis", email: "carol@example.com", age: 25 },
      { name: "David Wilson", email: "david@example.com", age: 42 },
      { name: "Eve Brown", email: "eve@example.com", age: 31 }
    ];

    for (const user of sampleUsers) {
      await connection.execute({
        sql: "INSERT INTO users (name, email, age) VALUES (?, ?, ?)",
        params: [user.name, user.email, user.age]
      });
    }

    console.log("Querying all users from the database...");

    // Query all users
    const allUsers = await connection.execute({
      sql: "SELECT * FROM users ORDER BY created_at"
    });

    console.log("All users:");
    console.log(JSON.stringify(allUsers, null, 2));

    console.log("Querying users older than 30...");

    // Query users with a condition
    const olderUsers = await connection.execute({
      sql: "SELECT name, email, age FROM users WHERE age > ? ORDER BY age DESC",
      params: [30]
    });

    console.log("Users older than 30:");
    console.log(JSON.stringify(olderUsers, null, 2));

    // Download the sqlite database
    const download = await agentdb.getDownloadUrl(newDatabaseToken, "users_db", "sqlite");
    const database = await fetch(download.downloadUrl).then((res) => res.arrayBuffer());
    // Write the database to a file using node
    fs.writeFileSync(download.fileName, Buffer.from(database));

    console.log("Database operations completed successfully!");
  } catch (error) {
    console.error("Error during database operations:", error);
  }
};

main();
CREATE A DATABASE

FEATURES

Vector Search

Built-in vector search with sqlite-vec extension. Store embeddings, perform semantic search, and build RAG systems.

SQLite + Duck DB

Maintain conversation state, user preferences, and context across sessions. Each conversation gets isolated storage.

No Setup Time

Databases are ready instantly. Serverless architecture scales from zero to millions of requests.

AGENT DB Database as Service
Provisioning Instantly allocate 1 or 1,000,000 databases without setup or configuration, just a unique ID. Manual setup process involving database creation, user management, certificate issuance, and network or firewall configuration.
Configuration & management Batteries included. Connect as many clients as needed to as many databases as required, all over HTTP. Vector search for LLMs works out of the box. Need to install and manage SQL extensions, configure appropriate measures like row-level security, and preallocate capacity to handle your clients.
Cost Pay only for data stored and queries executed, with no server or configuration costs. Manual scaling requires forecasting usage ahead of time and knowledge of provider-specific tiers to find optimal price point. If underutilized you are billed for idle compute or if overutilized your performance is harmed.
Portability Download any database as a single file that can be run anywhere, instantly. Exporting requires complex scripts or full database dumps, often at the expense of performance or uptime. Must spin up an entire database server to make use of the data.
Data provenance True tenant isolation, user data is separated at the filesystem level; malicious queries cannot cross user boundaries. Unified database makes finding and acting on data regulation more complex. Security holes risk widespread data breaches.
CREATE A DATABASE

SIMPLE
TRANSPARENT
PRICING

Starter

$0

  • 1M agent requests/month
  • SQLite & DuckDB support
  • Vector search with sqlite-vec
  • Unlimited databases
  • Community support
  • 1GB storage
Get Starter

Pro

$29/month

  • 10M requests/month
  • All Starter features
  • Priority support
  • Advanced monitoring
  • Custom domains
  • 100GB storage
Choose Pro

Enterprise

CUSTOM

  • Unlimited requests
  • All Pro features
  • Dedicated support
  • SLA guarantees
  • Custom integrations
  • Unlimited storage
Contact Sales
CREATE A DATABASE