After you configure your MCP (Model Context Protocol) client with the appropriate MCP server endpoint and valid user credentials, you unlock a set of MCP tools that let a third-party agent (such as Microsoft Copilot) interact directly with Automation Anywheres automation repository and runtime.

These tools enable third-party AI agents to discover, run, and retrieve results from Automation Anywhere automations through secure agent connections.

Tool Base License Enterprise License (PRE enabled)
DiscoverAutomation Not available Available
RunAutomation Not available Available
GetAutomationResult Available Available
Note:

The DiscoverAutomation and RunAutomation tools require an Enterprise license with Process Reasoning Engine (PRE) integration enabled. PRE integration is currently supported in US and Europe regions only.

DiscoverAutomation

The DiscoverAutomation tool discovers automations which are accessible to that user with the Run/Schedule permission in the automation repository. This tool enables AI agents to dynamically determine which automation to run based on user intent.

How it works:

  • The tool evaluates the user’s request.
  • It searches the automations and returns only those automations that the user is authorized to access. From the list of matching automations:
    • If nine or fewer automations are available, all are displayed.
    • If more than nine automations are available, only the top nine automations are returned.
    • The top nine are determined using internal relevance and ranking criteria. If an automation does not meet the ranking criteria, it might not appear in the default results.
To retrieve specific automations that are not shown in the default list, use a tool-specific query format:
Discover automations to <specific task>
For example: 
Discover automations to create invoice in SAP
Discover automations to extract data from Salesforce

Using specific task-based queries improves matching accuracy and increases the likelihood that the intended automation appears in the results.

Best practices:

  • Provide descriptive summaries for automations.
  • Limit unnecessary automations in a single agent scope when possible.

RunAutomation

The RunAutomation tool executes an automation that has been identified through DiscoverAutomation or explicitly specified by the agent.

It can run:
  • Bots (automations)
  • Processes
  • API Tasks

Execution is initiated from a third-party AI agent through the MCP server.

How it works:

  1. The automation is identified.
  2. Required input parameters are passed.
  3. The MCP server validates permissions.
  4. The Control Room schedules and runs the automation.
  5. A job ID or execution reference is returned. This reference is required to retrieve the execution status or results using GetAutomationResult.

The automation runs in the configured runtime environment and Role-based access control (RBAC) policies are enforced. Input variables must match the automation’s expected parameters. All execution events are logged in the Control Room audit logs.

Best practices:

  • Incorrect or missing input parameters might cause automation failure.
  • The automation must be published and enabled as an agent connection.
  • The user must have permission to run the automation.
  • Network and credential configuration must be valid.

If execution fails, use GetAutomationResult to retrieve error details.

GetAutomationResult

The GetAutomationResult tool retrieves the status or output of an automation that was started using RunAutomation.

This tool can be used with either:
  • A Base license
  • An Enterprise license with PRE enabled (US and Europe regions)

How it works:

  • The tool accepts a job ID or execution reference.
  • It queries the Control Room for current execution status which includes these possible status responses:
    • Running/InProgress: The automation is currently executing.
      Note: If the status is Running/InProgress, the tool can be called again later to retrieve the updated status or final result.
    • Completed: The automation finished successfully. When the status is Completed, the tool returns any configured output variables defined in the automation. Output might include:
      • Text values
      • JSON objects
      • Processed transaction results
      • Other defined return parameters
    • Failed: The automation execution failed. If the status is Failed, the response includes relevant error details and logs to assist with troubleshooting.

Security, governance, and tool interactions

All MCP tool operations:

  • Enforce Control Room authentication
  • Follow RBAC permissions
  • Maintain audit logging
  • Respect automation-level access controls

Only authorized users can discover, run, or retrieve automation results.

A typical tool interaction flow follows this sequence:

  1. DiscoverAutomation
  2. RunAutomation
  3. GetAutomationResult
This flow allows the AI agent to:
  • Identify the correct automaton
  • Execute it securely
  • Retrieve the final result

Troubleshooting MCP tools

Automations do not appear in DiscoverAutomation
  • Confirm user permissions.
  • Ensure automation is published.
  • Use a more specific tool query.
  • Check whether more than nine automations exist (top-nine limit).
Automation runs but no output is returned
  • Verify output variables are configured correctly.
  • Ensure output variables are mapped properly.
  • Confirm GetAutomationResult is called with the correct job ID.
RunAutomation fails
  • Validate input parameter names and types.
  • Confirm runtime environment availability.
  • Review Control Room audit logs.
  • Check PRE integration status (Enterprise license only).