# writerslogic/scrivener-mcp [Health: Active]

**Category:** 🧠 Knowledge & Memory  
**Repository:** https://github.com/writerslogic/scrivener-mcp  
**GitHub Stars:** 52  
**Views:** 16  
**Installs:** 12  
**Upvotes:** 6  
**Directory Page:** https://allmcps.com/mcp/writerslogic-scrivener-mcp

## Description
Connect Scrivener 3 writing projects to Claude and other AI assistants. 47 tools for document management, writing analysis, semantic search, character/plot memory, and content enhancement. Progressive skill loading, relationship engine with HMS triplets, and JS fallback for offline semantic search. npm i -g scrivener-mcp

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "scrivener-mcp": {
    "command": "npx",
    "args": ["-y","scrivener-setup"]
  }
}
```

## Documentation

## What writerslogic/scrivener-mcp MCP server does

The writerslogic/scrivener-mcp MCP server gives an AI client access to Scrivener 3 writing projects. An assistant can open a project, inspect and modify documents, search content, and perform writing-focused analysis. The repository describes 53 tools covering document management, analysis, content enhancement, offline semantic search, and character and plot tracking.

Supported project types include novels, screenplays, and manuscripts. The server is intended for Scrivener 3 on macOS, Windows, and Linux. It is compatible with Claude Desktop, Claude Code, VS Code integrations such as Copilot or Continue, Cursor, and other MCP-compatible clients.

## How it works

The server runs locally and communicates with MCP clients over stdio. After installation, the client can be configured to launch the `scrivener-mcp` executable. The assistant is then told which Scrivener project to open and can operate on the project through its available tools.

Core functionality does not require an external AI service. Deterministic analysis, document operations, keyword search, and project memory are available without an API key. The project also includes progressive skill loading and a relationship system based on Holographic Memory System data, described in the source material as HMS triplets.

Semantic indexing and similarity scoring use the local Holographic Memory System rather than a separate embedding API. The current semantic search flow can use the configured chat provider to interpret queries and explain results. When supported by the MCP client, some chat-based functions can instead use the client's sampling capability.

## Setup and configuration

Install the package globally with npm, or run it directly with `npx`. The repository also documents Homebrew, Docker, GitHub-based installation, and an interactive `scrivener-setup` command for detecting and configuring Claude Desktop, Claude Code, and Cursor. Other MCP clients can be pointed at `npx scrivener-mcp` as a stdio server.

The writerslogic/scrivener-mcp MCP server does not require an API key for its core features. AI-assisted analysis, generation, enhancement, and parts of semantic search can use Anthropic, OpenAI, or OpenRouter. Provider selection can be controlled with `AI_PROVIDER`, and OpenRouter model selection with `OPENROUTER_MODEL`. Keys may be supplied through environment variables, supported `.env` files, key files, or macOS Keychain entries.

If multiple providers are configured, the server uses Claude for chat and generation by default and can retry with another configured provider after an account-level failure. The README also documents provider-specific variables including `ANTHROPIC_API_KEY`, `OPENAI_API_KEY`, and `OPENROUTER_API_KEY`.

## Tools and capabilities

Available functionality includes:

- Opening and managing Scrivener project documents
- Reading and editing manuscript content
- Running deterministic writing analysis
- Enhancing or generating content with a configured AI provider
- Searching by keywords and using semantic search
- Tracking characters and plot information
- Maintaining project memory and relationships
- Loading skills progressively as needed

The exact tool count is reported as 53 in the repository README; the supplied listing description reports 47, so the count may vary by version or source.

## Limitations and notes

AI-powered features depend on a configured Anthropic, OpenAI, or OpenRouter key unless the MCP client provides supported sampling. Core operations remain available without those credentials. The writerslogic/scrivener-mcp MCP server works with Scrivener 3 projects, so projects from other writing applications are outside the stated scope.

The documented Docker example mounts a local projects directory into the container. For clients other than those auto-detected by `scrivener-setup`, configuration must be supplied manually as a stdio server. The provided material does not identify a software license by name, so licensing should be verified in the repository before redistribution.

_Full upstream README: https://allmcps.com/mcp/writerslogic-scrivener-mcp/readme_