A desktop spellbook for LLMs that installs MCP servers for you

What it does

Tome is a desktop chat client for local and remote LLMs—Ollama, Gemini, OpenAI, and OpenAI-compatible endpoints—that wraps the messy parts of the Model Context Protocol in a graphical interface. It lets you install, toggle, and run MCP servers, which give your LLM access to files, APIs, and search engines, without editing JSON configs or wrangling Node and Python manually. You can also schedule prompts to run hourly or daily, and an experimental Telegram relay lets you message your setup from your phone.

The interesting bit

Most MCP clients assume you enjoy debugging uvx paths in a terminal; Tome assumes you don’t. It bundles support for npm, uvx, node, and python servers internally and plugs into the Smithery.ai registry so you can add capabilities the way you’d install a browser extension. That “no-code” positioning is rare in a space still dominated by config.json tutorials.

Key highlights

• Runs local models (Ollama, LM Studio, Cortex) or cloud APIs (Gemini, OpenAI) interchangeably
• One-click MCP server installation via Smithery.ai registry; built-in support for npm, uvx, node, and python servers
• Scheduled tasks: set prompts to run hourly or daily with any model or MCP combo
• Experimental Telegram Relay lets you message your LLM setup from anywhere
• Native UI for tool calls and reasoning model output, so you aren’t squinting at raw JSON

Caveats

• Self-described “Technical Preview” with rough edges; Linux support is not yet available
• The Telegram relay and future app-builder features are experimental or still on the roadmap

Verdict

Worth downloading if you want LLM agents but find the usual MCP setup tedious, or if you’re helping a non-technical friend get started with local AI. Hardcore CLI enthusiasts who already maintain fifty mcp-config.json files will find it superfluous.