Logic Server

Agent reasoning, memory, and token-optimized context for AI applications.

Logic Server is an MCP server published by ai.nocturnus under the canonical name ai.nocturnus/logic-server. It is designed to supply AI agents with reasoning support, persistent memory, and token-optimized context management within AI applications. The server is distributed as the npm package nocturnusai-mcp at version 0.3.11 and communicates over the stdio transport, meaning the agent host process spawns and talks to it directly over standard input/output.

Configuration is handled through four environment variables: NOCTURNUSAI_URL sets the endpoint of the backing Nocturnus AI service, NOCTURNUSAI_API_KEY authenticates the connection, NOCTURNUSAI_DATABASE specifies which database the server should target, and NOCTURNUSAI_TENANT scopes the session to a particular tenant. All four must be provided for the server to function.

The server’s stated purpose is to give agents a structured way to reason over problems, store and retrieve memory across interactions, and manage context in a token-efficient manner — concerns that are relevant when building long-running or multi-step AI workflows. No tool list is exposed in the registry record, so the exact callable surface cannot be described here.

This server is not a general-purpose Postgres interface and does not expose SQL query tools. It requires an active Nocturnus AI backend reachable at the configured URL.

Use cases

  • Provide an AI agent with persistent memory across multiple conversation turns
  • Manage and compress context windows to reduce token consumption in long workflows
  • Scope agent memory and reasoning to a specific tenant in a multi-tenant AI application
  • Connect an MCP-compatible agent host to a Nocturnus AI reasoning backend

When to use it

  • Building multi-step AI agent workflows that need memory across turns
  • Working with a Nocturnus AI backend and needing an MCP integration layer
  • Token budgets are a concern and context optimization is required
  • The deployment environment supports stdio-transport MCP servers

When not to use it

  • Looking for a general-purpose Postgres or SQL MCP server
  • No access to a running Nocturnus AI backend at a known URL
  • The agent host requires a network transport (HTTP/SSE) rather than stdio
  • Needing a server with a documented, inspectable tool list before adoption