oneshot

MCP.Pizza Chef: Destiner

Oneshot is an early prototype MCP client developed by Anthropic for macOS, designed to integrate with the Model Context Protocol ecosystem. It enables users to bring their own API keys, discover available tools automatically, and install them with a single click. While currently in an early development stage, it offers a streamlined interface for managing MCP tools and interactions on macOS. Users should note that chat history and tool configurations may be lost in future updates. The client requires local dependencies such as bun and Python's uv, with plans to bundle these in later versions for easier setup.

Use This MCP client To

Manage MCP tools on macOS with a native client Easily configure and use your own API keys Discover and install MCP tools with one click Prototype MCP client interactions on macOS Run and test MCP API server and UI locally

README

oneshot

Anthropic's MCP client for macOS.

Warning

This is an early prototype. Expect the chat history and the tool configuration to be lost with future updates.

Features

Bring your own API key
Built-in tool discovery
One-click tool installation

Set up

To install dependencies:

bun install

Note

You also need bun and Python's uv installed locally. Later versions of the app will bundle those dependencies automatically.

To run:

Run the API server:

bun run server:dev

Run the UI:

bun run app:dev

oneshot FAQ

How do I install dependencies for oneshot?

You need to run 'bun install' and have bun and Python's uv installed locally.

Can I use my own API key with oneshot?

Yes, oneshot allows you to bring your own API key for authentication.

Will my chat history and tool configuration be saved?

Currently, chat history and tool configurations may be lost with future updates as it is an early prototype.

How do I start the oneshot client?

First run the API server with 'bun run server:dev', then start the UI with 'bun run app:dev'.

Is oneshot compatible with other LLM providers besides Anthropic?

While designed for Anthropic, MCP clients generally support multiple providers like OpenAI, Claude, and Gemini through the protocol.

Are there plans to simplify dependency management?

Yes, future versions aim to bundle bun and Python's uv automatically for easier installation.

What platforms does oneshot support?

Oneshot is currently developed specifically for macOS.

Is oneshot production-ready?

No, it is an early prototype and may have unstable features and data loss risks.