Give Your AI Agents the Context They Need Without Giving up Control of Your Data

EDB Postgres AI Agent Factory, now in AWS Marketplace, runs AI work inside Postgres, so your data stays in the database and your existing controls govern your agents, too.

Agents are poised to change how work gets done. But without semantic and business context, they hallucinate, and an agent that makes things up can’t be trusted with an enterprise’s customers, data, or decisions. Gartner expects organizations that prioritize semantics in AI-ready data to improve agentic accuracy by up to 80%. Getting that context to agents reliably is where most teams get stuck, because their agents don’t understand business data, can’t find the right documents at scale, and require sensitive data to leave the enterprise.

A traditional report runs the same pre-written query every time. An agent starts with only a question. It sends that question to an LLM, which translates it into SQL. But the LLM doesn’t know how the data is organized or what the business means by “revenue” or “active customer,” so it guesses at tables, columns, and joins. Often the query fails. When it does run, it returns a wrong answer with high confidence.

A second problem is that much of what agents need isn’t in the database at all. It lives in documents, images, and other unstructured content in separate data stores such as SharePoint. In a September 2026 VentureBeat survey, 41% of enterprises running AI agents said retrieval over documents is their agents’ main source of business context. But finding the right file across massive document stores is difficult and expensive at enterprise scale.

Then there’s the issue of security. Building AI today usually means sending sensitive data over the network to external providers for inference, which exposes it to three risks: Bad actors can read it in transit; the inference service may hold it in temporary storage that could be breached; and the provider may use it to train future models, potentially exposing it to other users. For organizations that value data sovereignty and residency, every transmission that leaves their environment adds to that risk.

EDB Postgres AI solves all three of these challenges with its Agent Factory capabilities, which bring everything agents need—context, enterprise search, and a model and agent runtime—to the database. With this unique approach, data no longer leaves the enterprise environment, making it governed and sovereign by default. 

The semantic layer in EDB Postgres AI enables agents with business context. It captures how the data is organized and what terms such as revenue or active customer actually mean. So when someone asks, “How many active customers did we lose last quarter?” the agent writes the right query instead of guessing. 

Advanced vector search finds the right information across document stores at enterprise scale. In McKnight Consulting Group benchmarks at 50 million vectors, EDB Postgres AI returned results in 50 ms at the median, compared with 1,083 ms for MongoDB Atlas and 4,023 ms for Databricks, and posted the highest recall of any platform tested. It also led every other tested Postgres provider on speed and accuracy, and on cost once performance is factored in.

An in-database runtime runs agents and models right next to the data, so customer records, financial information, and other sensitive content never have to travel to an outside provider. For heavier workloads, organizations can run agents outside the database and connect them securely through EDB’s MCP Server, gaining the scale they need without giving up control of their data.

These agentic AI capabilities are available directly in EDB Postgres AI, which can run anywhere, including on premises or in air-gapped environments. And now it’s also available in AWS Marketplace, packaged and optimized for customers who want it inside their own AWS environment. 

One click deploys it as a single Amazon Machine Image and CloudFormation template in the customer's VPC. CloudFormation builds the networking, storage, and access permissions around the image, so the setup work is already done. EDB Postgres AI Agent Factory integrates with AWS AI services, so teams can build end-to-end agentic workflows. Amazon Bedrock plans the work and writes the queries, Amazon Bedrock AgentCore hosts the agent and manages identity and access, and EDB Postgres AI provides a governed semantic layer inside Postgres that the agent reaches through its built-in MCP endpoint.

Another valuable feature of EDB Postgres AI Agent Factory is that the Postgres database it deploys is Oracle compatible. Getting ready for agents often means modernizing the infrastructure, and for many enterprises that includes expensive legacy infrastructure on Oracle. Migrating off it competes with the AI initiatives for the same people and budget, so each project slows the other. 

But because EDB Postgres AI is Oracle-compatible, Oracle applications can move to AWS with minimal code changes. The same Postgres database that hosts the migrated applications also gives agents context, enterprise search, and a model and agent runtime. Teams get an AI-ready foundation from one move instead of two projects that compete for resources.

Agents give good answers when they can reach live data and understand what it means, and that data already sits in the enterprise database. EDB Postgres AI Agent Factory puts the AI work right beside it, and AWS customers can deploy it in their own account straight from AWS Marketplace.

Learn more about EDB Postgres AI Agent Factory for AWS on AWS Blogs.

Get started today with EDB Postgres AI Agent Factory for AWS on the AWS Marketplace.