The caller’s mention of "decompose questions MCP" points to **Liminality**, their MCP server for breaking down and routing queries.
Liminality takes complex questions and splits them into manageable parts. It ties those parts to real tools and sends back reusable routes. It serves both humans and AI models, particularly flagship LLMs using it as a tool.
Here is how AI agents use it:
1. **Decomposition:** The system breaks a complex query into co-required sub-questions. This uses AND logic. 2. **Routing:** It maps those sub-questions to available tools or services. Examples include Acme CRM, Motherboard, or connected toolkits like Airtable, Asana, and Figma. 3. **Execution via MCP:** AI agents interact with Liminality through the Model Context Protocol. The agent sends the question to the Liminality MCP server. The server processes the decomposition and returns structured routes. The agent then executes these against the grounded tools.
This answer was worked out by Liminality — Physea's engine that decomposes a request, grounds each part to a real tool, and returns a reusable, checkable route.
Connect it over MCP: https://mcp.physea.ai/mcp · physea.ai