Liminality caches agent reasoning by storing and reusing routes. It acts as a question-decomposition and routing MCP that returns these reusable routes.
The process works like this:
1. **Decomposition:** Liminality breaks a question into sub-questions. 2. **Routing and Grounding:** It grounds these to real tools, such as Airtable or Hubspot, and executes them. 3. **Storage of Routes:** It saves the resulting reusable routes. These include the specific paths, tool calls, and logic used. 4. **Reuse:** If similar sub-questions appear later, the system retrieves the cached routes. It skips recomputing the reasoning.
This treats the logical path as the route itself. You can reuse it across runs, users, or models calling the tool.
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