YugabyteDB AMP
(Agentic Multitenant Postgres)

Serverless Multitenant PostgreSQL
Run by Agents

Agents are bursty and frequently idle, but you’re billed for capacity whether you use it or not.

YugabyteDB AMP (Agentic Multitenant Postgres) is serverless, scale-to-zero, multitenant PostgreSQL that can be operated by agents. AMP’s built-in agents handle setup, migration, tuning, and integrations. Zero-copy branching lets you test changes safely.

Teams can spin up hundreds of isolated databases in seconds and group them into shared pools, each with its own quotas and management policies. As workload usage and business criticality grow, you can move to YugabyteDB YSQL on a clear, no-migration upgrade path, with no app rewrite or replatforming.

Find Out More

YugabyteDB AMP: Agentic MultiTenant PostgreSQL

One Platform for Fleets of Databases

YugabyteDB AMP’s serverless multitenancy packs hundreds of isolated PostgreSQL databases onto shared infrastructure, organized into shared pools with their own quotas and policies.

Instant Isolated Databases

Spin up PostgreSQL clusters with an agent, MCP, the API, or the UI, and branch onto independent compute with zero-copy thin cloning. Shared pools set quotas and management policies for each team or project.

Only Pay For
What Runs

Usage-based pricing only charges for what runs. An idle database scales to zero, so small and early-stage workloads don't pay for idle capacity.

Grow Without
Rewrites

Workloads can start small and move to YugabyteDB YSQL on a no-migration path, with no app rewrite or replatform when needed. Each environment's storage layer can run across multiple zones, regions, or clouds to deliver the resilience you need.

Run by Agents

Day-to-day database tasks are handled by built-in agents that share context and maintain a full action decision trace. Dedicated agents configure setup, drive migrations, troubleshoot, tune performance, and manage external integrations.

How YugabyteDB AMP Works

Step 1: Provision a Database Per Agent

YugabyteDB AMP’s Serverless Multitenancy efficiently packs hundreds of small agent workloads onto shared, distributed infrastructure. Every database is standard PostgreSQL, with vector search and graph queries, supporting agent workloads that need more than relational data.

Step 2: Let Agents Operate Themselves

Day-to-day database work is handled by four built-in agents. YugabyteDB Architect composes the setup, YugabyteDB Voyager drives migrations (including Oracle, SQL Server, and MongoDB), YugabyteDB Perf Advisor troubleshoots and tunes performance, and YugabyteDB Nexus manages external integrations like connectors, event streams, and BI tools.

Step 3: Keep Costs Predictable

Resource Governance caps CPU per workload now, with RAM, storage, and network caps on the roadmap. A runaway query can’t consume another workload’s CPU. Spin up hundreds of databases with built-in quotas and manage them as a single fleet.

Step 4: Scale When It Matters

As an agent grows in use and importance, move your workload to YugabyteDB YSQL on a no-migration upgrade path, with no need for migration or rebuilding. One platform and one commercial relationship across your entire agentic AI journey.

Features and Capabilities

Serverless Multitenancy

True scale-to-zero for bursty, frequently idle agent workloads. Consumption-based pricing with no pass-through fees.

Resource Governance

Per-workload CPU caps available now, with RAM, storage, and network caps coming soon. Prevent one workload from crowding out another, while still allowing bursts to full hardware capacity.

Seamless Upgrade Path

A clear, no-migration upgrade path to YugabyteDB YSQL when a workload grows in use and business criticality, with no app rewrite.

Fleet
Management

Batch-create databases with built-in quotas. Manage hundreds of agent workloads as a single fleet.

Agent-Operable
Lifecycle

Four built-in agents handle setup, migration, tuning, and integrations, with zero-copy branching for safe testing. Every agent action is recorded in a full decision trace.

Connection Manager

A built-in server-side pooler that multiplexes thousands of client connections onto a small backend pool: no external PgBouncer required.

YugabyteDB MCP 2.0

Native MCP server with OIDC-based authentication and write-capable tools that stay disabled until an operator enables them, so agents reach production data under explicit control.

Deploy Anywhere

Deploy YugabyteDB AMP in your own cloud account or on-prem Kubernetes, and manage it via the YugabyteDB cloud control plane.

“By 2028, AI agent ecosystems will enable networks of specialized agents to dynamically collaborate across multiple applications, allowing users to achieve goals without interacting with each application individually.”
(Gartner, August 2025)

Optimize Slow Database Queries With AMP

Watch this short demo to discover how to optimize slow database queries using an autonomous agent combined with zero-copy agentic branching technology.

Problem Identification: A slow query in a production environment is identified and passed to an agent for resolution without requiring extensive human guidance.
Agent Analysis: Utilizing native connections to database schema and workload patterns, the agent analyzes the issue and proposes multiple potential solutions.
Isolated Branching: To avoid impacting production performance or requiring risky manual changes, the agent creates three separate, isolated branches of the production database.