holy-moly-mcp-remote

# holy-moly-mcp-remote Lightweight stdio [MCP](https://modelcontextprotocol.io/) client that proxies conversion tool calls to an already-running [`holy-moly-web`](https://github.com/the78mole/holy-moly-mcp) deployment over HTTP, instead of loading faster-whisper / marker-pdf / WeasyPrint locally. Only depends on `fastmcp` and `httpx`. ## Install ```bash pip install holy-moly-mcp-remote # or uvx holy-moly-mcp-remote ``` ## Configuration | Env var | Description | |---|---| | `HOLY_MOLY_API_URL` | Base URL of the remote `holy-moly-web` FastAPI backend | | `HOLY_MOLY_API_TOKEN` | Bearer token for that backend (static API token or a token from `POST /api/v1/auth/login`); leave empty if the deployment has no auth. Sent only to the API, never to third-party download URLs. | | `HOLY_MOLY_DEFAULT_WHISPER_MODEL` | Default speech-to-text model to request (optional; per-call `model` argument overrides it) | | `HOLY_MOLY_DEFAULT_TTS_MODEL` | Default Kokoro TTS model to request (optional; per-call `model` argument overrides it) | ## Tools - `convert_speech_to_text` — transcribe a local file or remote URL (`mode=plain|minutes`, optional `language`, optional `model`) - `convert_pdf_to_markdown` — convert a local PDF or remote URL to Markdown (marker-pdf) - `convert_html_to_pdf` — render HTML to PDF (WeasyPrint) and save it locally - `list_tts_voices` — list available text-to-speech voices - `convert_text_to_speech` — synthesise speech to a local WAV file (optional `model` for Kokoro voices) ## VS Code MCP configuration ```json { "servers": { "holy-moly-remote": { "command": "uvx", "args": ["holy-moly-mcp-remote"], "env": { "HOLY_MOLY_API_URL": "http://your-holy-moly-web-host:8080", "HOLY_MOLY_API_TOKEN": "hm_..." } } } } ``` See the [main repository](https://github.com/the78mole/holy-moly-mcp) for the full project, including the local in-process MCP server and the FastAPI/Vue web stack this client talks to.