---
nav: wcp_docs
parent_url: /wcp_docs/
title: Agent-Ready MCP Tools
layout: subsection
---

Agent-Ready Model Context Protocol (MCP) tools are purpose-built interfaces made specifically for autonomous Al consumption. These tools pass a strict validation framework to ensure agent-readiness, secured by default with Delegate Authorization safeguards and large language models (LLMs), and are optimized with flattened schemas, rich metadata, and automated error remediation. Flat schemas and self-describing tool interfaces minimize hallucinations and reduce latency so agents can operate at enterprise scale.

Workday exposes numerous Agent-Ready MCP tools on the MCP Explorer that cover Workforce, Financials, Talent, Payroll, Recruiting, and more. These Agent-Ready MCP tools are callable by any agent, on any platform, that has been granted access through Agent System of Record (ASOR). Workday's MCP tools are designed to be discovered and invoked at runtime by Al agents that read schemas and descriptions on each interaction.

You can call MCP tools in:

-   Workday custom agents. MCP tool discovery and registration are handled automatically when you Save and Deploy. See [Build and Deploy Custom Agents](/wcp_docs/GUID-c2e04f80-557a-4968-8957-9f1210156908-enHYPHENus.html).
-   External platforms such as Claude, Cursor, Copilot Studio, Google Agentspace, and others. You can only call Workday MCP tools once you register your agent in ASOR, configure which tools it can call, and connect to the Workday MCP server. See [Connect an External Agent with Workday MCP](/wcp_docs/GUID-03605c39-71d3-4314-bf71-97f274df457c-enHYPHENus.html).


Workday Agent-Ready MCP Tools are built on the industry standard for agent communication: Model Context Protocol JSON-RPC 2.0 spec. Your agent connects to a single endpoint and calls tools using the MCP protocol. The LLM handles tool selection and call syntax automatically.

<b>Note:</b> Agent-Ready MCP tools are for Al Agent use only, not for hard-coded integrations. Because LLMs are inherently non-deterministic, expect execution behavior and results to vary across different runs, clients, prompts, and models.

## MCP endpoint

```
https://{agentGatewayRegionEndpoint}/{tenantName}/mcp
```

Example:

```
https://us.agent.workday.com/{tenantName}/mcp
```

## Security

Agent-Ready MCP tool authentication uses only OAuth 2.1 with a Proof Key for Code Exchange (PKCE) token. Most agent platforms and API clients (Bruno, Postman) handle the PKCE OAuth 2.0 flow natively.

MCP calls use the Agent System User (ASU) OAuth flow at runtime. You must obtain a Workday-specific ASU token when creating and registering your agent definition in Agent System of Record (ASOR). This token allows every call to be authorized against both the agent and the user without asking the user to log in again.

Agent definitions in the Agent System of Record are made up of skills consisting of multiple Agent-Ready MCP Tools. You can configure security on the skill (a grouping of tools) and on the individual tool level. With ASU security in delegate mode, agent permissions are never elevated beyond the scope of the user the agent is acting on-behalf of.

## Versioning

Workday delivers MCP Tools as a single, continuous surface without traditional versioning. To optimize how agents interact with the tools, you can expect continuous unversioned updates such as schema restructuring, tool renaming, and description rewrites. The MCP Explorer updates weekly on Tuesdays. You can export a CSV file to view a change log of all tool additions, removals, and changes in the MCP Explorer list.

For more details, see [Workday Agent-Ready MCP Tools: Developer Guidelines & Change Policy](/wcp_docs/GUID-04498a6d-10ba-4a41-a917-9fa668641b8d-enHYPHENus.html).

## Limits

Workday enforces these service protections for Agent-Ready MCP Tool calls:

-   Active Request Throttling

    -   Concurrent Request Limit = 2,000 requests

-   Rate Limiting

    -   Refresh Rate = 500 requests
    -   Burst Limit = 500 requests



