AI NewsInfrastructureAnnouncement
Workday opens HR and finance records to outside AI agents through MCP
Workday says its new Trusted Agent Platform lets outside AI agents read and act on HR and finance records through Model Context Protocol tools, with security and audit rules travelling on every request.
Image: Workday via The New Stack
Why it mattersA developer building an agent that touches employee or payroll data has had to copy that data out of Workday first, and this is the first time Workday says the agent can read it where it lives, under Workday's own permissions.
The data an HR agent needs sits inside a system of record that nobody writes code against from outside. Workday says it has now opened that gate.
At its Workday Rising event, the company announced the Trusted Agent Platform, which it describes as a way for developers to build AI agents that read and act on live Workday HR and finance data using the tools and models they already use. The announcement, written by Gabe Monroy, says the core capabilities are available today, with more to arrive late 2026 and early 2027. A second post the same day, by Yasmeen Ahmad, introduces the Workday Data Cloud underneath it.
What is in the platform
The platform has five named parts. Agent-Ready Tools are Model Context Protocol servers that Workday says are compatible with Claude, GitHub Copilot and other agent stacks. The Workday Data Cloud gives SQL access to live business objects through Apache Iceberg and Trino federation, which Workday says means the agent does not need a copy of the data. The Workday Agent Harness is a runtime that Workday says enforces memory, guardrails, routing, model choice and audit traces. A Developer Agent turns a natural language prompt into a production agent, and Custom Agents are supported through Google ADK, LangChain and AWS frameworks.
What Workday says about security
Workday's quote on this is explicit: "Headless must be lawful. Security context and business rules have to travel with every single request." The company says deterministic grounding, native contextual security, and a full audit log sit under every agent action, and that the Data Cloud federates to Salesforce, Snowflake, Power BI, Databricks and BigQuery without copying the records across. One customer quote, from David Cavanaugh at Werner Enterprises: "It feels like magic because we can focus on insights and outcomes instead of worrying about compute."
What a developer can do with it now
The reader of this item is the person who was going to build an agent that answered a question like "what did we spend on contractors last quarter in Germany" or "who on this team is overdue for a performance review". Until now, that agent had to either call Workday's public REST API, which gives a small window on the data, or wait for a nightly extract into a data warehouse the agent could then query. Workday says the agent can now call an MCP server, read live tables under its caller's own permissions, and leave an audit trail that names a human.
What Workday has not said yet: an exact list of tables an MCP tool can read, what rate limits apply, whether the Data Cloud and the Trusted Agent Platform are included in a current Workday contract or priced separately, and whether a non-customer developer can call any of this. The post names release windows but no price. Anyone starting a build against this today should assume that part is still a conversation with their Workday rep.
Source
Introducing Workday's Trusted Agent Platform, Gabe Monroy, Workday. See also Headless, Open, and Secure: Meet The New Workday Data Cloud, Yasmeen Ahmad, Workday.
This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.
Get AI News in your inbox
New developer tools, model and agent releases, and how teams are actually using them to release software. Short, and only when there is something worth reading.
