Ataccama launches headless MDM for AI agents & apps

Ataccama has launched headless master data management in Ataccama ONE MDM 17.2, aimed at AI agents and business applications that need direct access to master data.

The update lets AI agents and external workflows read and write master data through Ataccama's MDM MCP server or REST API. Teams can use master data in agent tools such as Claude Code and Codex, as well as in business workflows that call MDM directly.

Master data management systems maintain a consistent version of core business entities such as customers, products and suppliers. Ataccama's move focuses on making those records available outside the traditional user interface, so automated systems can check, update and route work around the master data layer.

The release also includes a new user interface designed to support work carried out in a headless model. In practice, an automated process can create a task for a human data steward to review and approve, while keeping an audit trail of what has been done.

The announcement addresses two common issues in AI-driven workflows. One is the risk that an agent without access to a trusted customer or product record creates duplicate entries or misses relevant information. The other is the growing volume of stewardship work, in which staff review uncertain matches and record proposals by hand.

Under the new model, business process agents can query what Ataccama describes as golden records before taking action. Stewardship agents can also handle repetitive tasks, including reviewing matching proposals in bulk, proposing record merges and creating tasks for supervisors or approvers.

Teams can decide how much autonomy to give those systems. An agent's work can remain in draft form, be sent to a supervisor for approval or be published, depending on the controls an organisation puts in place.

The release also provides access to matching rules and cleansing codes through the MCP server. That allows a reviewer to ask an agent why two records have been proposed as a match before deciding whether to approve the recommendation.

Ataccama has also expanded its REST API to cover a broader set of master data management functions, giving customers an alternative route to the same headless model if they choose not to use MCP-based integrations.

The development reflects a wider shift in enterprise software as suppliers adapt products for AI agents rather than solely for human users working through screens and forms. In master data management, that raises questions about how trusted records are made available to automated processes while retaining human oversight for higher-risk decisions.

Ataccama cited Gartner research on the role of master data management in creating a trusted context layer for AI agents. The issue is particularly relevant in processes such as customer onboarding, where an automated system needs to know whether a customer already exists in order to avoid duplication and incomplete records.

Nickolaos Mikedis, Leader of Strategy and Data Trust, Global Customer Experience, iA Financial Group, described the operational backdrop for the technology in his organisation.

"With about a dozen source systems behind our golden record, from legacy on-premises platforms to modern cloud applications, Ataccama MDM processes many events a day and pushes trusted records to our consumers in real time, ensuring smooth analytical and operational work. The MDM MCP server is what excites us next. Agents that help with matching and automate low-level stewardship tasks will give our data stewards new efficiencies as we expand our Data Trust Hub across iA," said Mikedis.

Jessie Smith, Chief Product Officer at Ataccama, said the change reflects the way AI systems now interact with business data and operational tasks.

"For years, MDM did its most important work behind a screen, with stewards reviewing uncertain matches by hand. That model does not keep up when AI agents need golden records at the moment they act, and it breaks entirely when agents start handing work to other agents. Headless MDM puts trusted master data where those decisions happen and lets agents take on repetitive stewardship work, while people stay in control of the calls that need human judgment. The payoff is that agents can be trusted with more consequential work, because they are all acting on the same record," said Smith.