Workday gives external AI agents access to its HR and finance data
Each agent has the same access limits and approvals as the person that it works for.
On Tuesday, at its Workday Rising conference in Las Vegas, Workday made its Trusted Agent Platform headless. Developers can make agents for HR and finance in tools, for example Claude Code and Cursor. An MCP server, SQL access to live data and a runtime for agents are available to all customers. Workday reports that each agent follows the same limits as the person who uses it.
What is available
Workday reports that these parts are available:
- Agent-Ready Tools: an MCP server with 275 read tools and 54 tools that write or erase data
- Live Data Query: lets an agent send SQL queries to live Workday data
- Agent Harness: a runtime for all agents that sets limits for each agent
How access works
A person connects with the Workday credentials of that person. Workday reports that an agent cannot get data or tools that the account of the person cannot get.
The demo
Gabe Monroy, the CTO of Workday, showed an agent on an MCP server that the customer made. The agent gave job requisitions that the person must not see. It also moved the incorrect employee between teams with no approval.
The same agent used a copy of the data in Snowflake and gave the salary of the CTO.
What is next
Workday will add more parts at the start of 2027: Agent Context Service and Grounded Reasoning. The first part gives an agent only the data for its task. The other part checks each step of an agent against the policies of the customer before the agent does it.
Sources
Posted
