MCP Simple OpenAI Assistant

Name: MCP Simple OpenAI Assistant
Author: andybrandt

@andybrandt

リポジトリ

MCP Simple OpenAI Assistant について

MCP server that gives Claude ability to use OpenAI's GPTs assistants

基本情報

カテゴリ

AI とエージェント

ライセンス

MIT

ランタイム

python

トランスポート

stdio

公開者

andybrandt

設定

以下の設定を使って、このサーバーを MCP 対応クライアントに追加してください。

{
  "mcpServers": {
    "mcp-simple-openai-assistant": {
      "command": "npx",
      "args": [
        "-y",
        "@smithery/cli",
        "install",
        "mcp-simple-openai-assistant",
        "--client",
        "claude"
      ]
    }
  }
}

ツール

(Create OpenAI Assistant) - Create a new assistant with a name, instructions, and model.

(List OpenAI Assistants) - List all available assistants associated with your API key.

(Retrieve OpenAI Assistant) - Get detailed information about a specific assistant.

(Update OpenAI Assistant) - Modify an existing assistant's name, instructions, or model.

(Create New Assistant Thread) - Creates a new, persistent conversation thread with a user-defined name and description for easy identification and reuse. This is the recommended way to start a new conversation.

(List Managed Threads) - Lists all locally managed conversation threads from the database, showing their ID, name, description, and last used time.

(Delete Managed Thread) - Deletes a conversation thread from both OpenAI's servers and the local database.

(Ask Assistant in Thread and Stream Response) - The primary tool for conversation. Sends a message to an assistant within a thread and streams the response back in real-time.

概要

What is MCP Simple OpenAI Assistant?

MCP Simple OpenAI Assistant is a Model Context Protocol server that lets tools like Claude Desktop create, manage, and converse with OpenAI assistants. It handles the full lifecycle of assistants and conversation threads, with local SQLite persistence for thread metadata since the OpenAI API does not expose a thread listing endpoint.

How to use MCP Simple OpenAI Assistant?

Install via pip install mcp-simple-openai-assistant or using Smithery’s CLI. Set the OPENAI_API_KEY environment variable and configure your MCP client (e.g., Claude Desktop) to run python -m mcp_simple_openai_assistant. The primary workflow: use create_new_assistant_thread to start a named conversation, list_threads to find existing threads, and ask_assistant_in_thread to stream responses.

Key features of MCP Simple OpenAI Assistant

Create, list, retrieve, and update OpenAI assistants
Create, list, and delete persistent conversation threads
Stream real-time responses from ask_assistant_in_thread
Local SQLite database for thread management
Simple pip or Smithery installation

Use cases of MCP Simple OpenAI Assistant

An AI assistant (e.g., Claude) creating its own OpenAI assistant for delegated tasks
Building multi-turn conversations with named, reusable threads
Managing multiple assistants and threads across sessions without losing context
Integrating OpenAI assistant capabilities into any MCP-compatible host

FAQ from MCP Simple OpenAI Assistant

What dependencies does the server require?

Python 3.8+, an OpenAI API key, and internet access. The server is installed via pip and runs as a Python module.

How are threads stored and retrieved?

Thread IDs and metadata (name, description, last used time) are saved locally in an SQLite database. This allows listing and reusing threads even though the OpenAI API does not support listing them.

Why does `ask_assistant_in_thread` use streaming?

OpenAI assistant responses can take a long time. Streaming provides real-time progress updates to the client, avoiding timeouts and improving user experience.

What authentication does it use?

The server reads the OPENAI_API_KEY environment variable. No other authentication or transport is exposed beyond the MCP standard (stdio).

Are there any known limitations?

The README lists planned improvements: listing available models, fine‑tuning assistant parameters (temperature, top_p), reading full thread history, and file upload support. These are not yet implemented.