ainovelprompter

AI Novel Prompter provides an MCP server integration that lets Claude Desktop (and other MCP-enabled clients) access and manipulate your story data in real time. The server exposes your characters, locations, codex entries, rules, and sample chapters so you can query, retrieve, or reference them directly within conversations and prompts. Use the Prompt Generation tab to assemble AI-optimized prompts for writing tasks (e.g., Write Next Chapter, Revise Chapter) and benefit from live previews and token counting. The Prose Improvement tab applies customizable improvement prompts and routes revised text through configurable LLM providers, enabling iterative refinement of prose while preserving story consistency. Claude Desktop can query entities, search locations, and access related data during conversations, streamlining creative workflows without leaving your writing environment.

Prerequisites:

• Go toolchain installed and available in your PATH
• Claude Desktop installed (for MCP integration) if you plan to connect via Claude
• Basic familiarity with JSON config files

Installation steps:

1. Build the MCP server

   • Open a terminal in the project root and run:

   cd mcp go build -o ainovelprompter-mcp.exe ..\cmd\mcp-server\main.go

   This compiles the MCP server into a Windows executable (adjust the path if building on another OS).

2. Configure Claude Desktop to load the MCP server

   • Create or edit claude_desktop_config.json and add your MCP server entry, for example:

   { "mcpServers": { "ai-novel-prompter": { "command": "C:\path\to\ainovelprompter-mcp.exe", "args": [ "--data-dir", "C:\path\to\your\story\data", "--log-level", "INFO" ], "env": {} } } }

   • Save the file and restart Claude Desktop to load the MCP server.

3. Run in development or production mode (optional)

   • For development, run the MCP with your preferred data directory and log level as shown in the example.
   • For production, point data-dir to a stable data location and set log-level to INFO or DEBUG as needed.

Tips and notes:

• Data directory structure is stored under the specified --data-dir path; ensure proper read/write permissions.
• Logging verbosity can be adjusted with --log-level (DEBUG/INFO/WARN/ERROR) to aid troubleshooting.
• Validation enforces non-empty names and descriptions for all entities; empty or whitespace-only entries will be rejected.
• If Claude Desktop cannot connect, verify the executable path, arguments, and that the data directory contains the expected subfolders (characters, locations, codex, rules, etc.).
• After updating claude_desktop_config.json, always restart Claude Desktop to apply changes.