RavenDB launches Quill for AI agents on SQL databases

RavenDB has launched Quill, a context layer for SQL databases aimed at production AI agents. The product is designed to work with existing SQL systems without requiring data migration.

Quill connects directly to an organisation's SQL database and sits between AI tools and the source system. This lets companies keep the original database as the system of record while adding search, retrieval and agent functions on top of live operational data.

The launch targets a common problem for large organisations with critical business data in long-established SQL environments. Many have struggled to connect that data to AI systems without embarking on costly modernisation projects or building custom infrastructure for data access, governance and security.

Quill is available for PostgreSQL, SQL Server and MySQL environments. It can be deployed in the cloud or on premises, which may help organisations meet data residency or regulatory requirements.

A Gartner survey cited by RavenDB found that one in five AI initiatives fail and that only 28% deliver a positive return on investment. RavenDB linked those outcomes to the difficulty of integrating AI with operational systems and controlling how models interact with sensitive information.

Governance focus

The product is built on the idea that AI models should not have unrestricted access to source data. Instead, Quill sits between the model and the SQL database, allowing organisations to define what an agent can retrieve regardless of the permissions structure in the underlying database.

That approach is intended to reduce the bespoke work needed to make AI use cases acceptable in regulated or sensitive settings. RavenDB gave the example of a healthcare deployment in which an agent could answer a patient's question about an appointment while being blocked from prescription records.

Quill is also model-agnostic, allowing customers to use different AI models, switch providers or run systems on their own hardware. That may appeal to businesses trying to avoid dependence on a single model vendor while still building services on top of existing operational data.

Agents built on Quill support web chat, WhatsApp, Telegram, Slack and Discord. RavenDB said customers can deploy working agents in weeks rather than the 18 to 24 months it associates with a typical in-house build.

Oren Eini, Founder and Chief Executive Officer of RavenDB, said the challenge for companies often begins after a prototype has proved it can work. "Anyone can stand up an AI demo in an afternoon, but getting that demo into production with data pipelines, semantic search, security, governance, all the plumbing a small proof of concept doesn't need until it has to run at scale, is the hard part," Eini said.

He said RavenDB built Quill to avoid repeated custom development by enterprise teams. "Quill exists because we'd rather hand teams that plumbing already assembled than watch them rebuild the same project after project. You get access to the live data you need, decide the scope on day one, and change it as you go, instead of building everything from scratch," Eini said.

Early use

RavenDB said Quill has already been adopted by Albos Technologies and Holdings, which it described as an early user of the product.

Hagay Albo, Chief Executive Officer of Albos Technologies and Holdings, said the product reduced the infrastructure work needed before application development could begin. "With Quill, the plumbing was already there, so we spent our time building the actual feature," Albo said.

RavenDB is best known for its NoSQL document database and says it has more than 12,000 customers across 50 industries. With Quill, it is expanding beyond database software into tools intended to help customers use AI with data already held in traditional SQL systems.

The move reflects a broader shift in the enterprise software market, where suppliers are trying to position themselves not only as data storage vendors but also as intermediaries between operational systems and AI models. In that market, products that can enforce controls over what an agent can see may prove as important to adoption as the quality of the model answering the question.