Licensing overview

Edit online

IBM® Engineering AI Hub uses token licensing, which allows users and tools to share token entitlements. This model helps organizations optimize license usage across multiple IBM products. For more information, see IBM Common Licensing documentation.

IBM Engineering AI Hub employs a modular licensing framework that adapts to a variety of deployment models:

  • Engineering Lifecycle Management tools
  • Large Language Model (LLM)
  • IBM Engineering AI Hub

Engineering Lifecycle Management tools

Engineering Lifecycle Management (ELM) users are provisioned with licenses, typically in the form of floating token licenses. These licenses support concurrent usage across the engineering team, allowing flexible access to ELM tools based on demand.

Large Language Model (LLM)

IBM Engineering AI Hub provides large language model (LLM) capabilities with multiple deployment options:

  • watsonx.ai

    • Cloud deployment: Offered as a monthly subscription with additional usage-based pricing.

    • On-premises deployment: Requires licensed software and GPU hardware for operation.

  • AWS Bedrock

IBM Engineering AI Hub

IBM Engineering AI Hub uses token-based licensing from the same pool as other ELM applications, enabling dynamic usage of AI capabilities.

Each Engineering AI Hub license corresponds to 25 tokens managed by the IBM® Common Licensing Server. A license entitles agents to process one work unit block of 10,000 Work Units (WUs), with 25 tokens required to refresh work units.

When both Agent-to-Agent and MCP endpoints are enabled, 100 ELM tokens are required to start or run Engineering AI Hub. These tokens provide access to a shared pool of 20,000 Work Units (WUs), which are consumed by agents and MCP tools.

A Work Unit is the unit by which program usage is measured. It represents:

  • An installation of Engineering AI Hub
  • A request to an agent from a user or another agent
  • The use of MCP tools by a user or an agent during execution of a request, each calculated per 24-hour period. For more information, see Licensing terminology.

Agent Work Units are consumed per request. Each request to an agent, whether initiated by a user or another agent, consumes the corresponding Work Units for that agent.

MCP tool–related Work Units apply to each request of an MCP tool by a user, tool, or agent, with each MCP tool call consuming the corresponding Work Units for that tool. Each MCP tool is classified by category (Read, Link, or Write) and complexity level (Core, Basic, Standard, or Advanced). For information about token distribution and shared work unit pools for agents and MCP tools, refer to WU allocation page.

For information about token distribution and shared work unit pools for agents and MCP tools, see Licensing model.