ask-openai
Ask a direct question.
LABSCO SUMMARY
Assistants API management from a client that is not OpenAI's.
The useful part is the lifecycle: an assistant you are already talking to can create another one, attach the files it should search, and adjust its instructions — work that otherwise means the dashboard or a script. ask-openai is the smaller half, worth having when you want a second model's answer in the same conversation.
What it is
A bridge between MCP and the OpenAI API. It exposes a direct chat completion alongside the Assistants API — create, list, update and delete assistants, upload and manage the files they search over, and turn file_search on or off — so an assistant can build and maintain other assistants.
What you get
- ask-openai — put a question straight to a chosen model, with temperature and max_tokens
- list-assistants and retrieve-assistant — the assistants on the account, and one of them by ID
- create-assistant — name, instructions, model, temperature, attached file IDs, and whether file_search is enabled
- update-assistant and delete-assistant — change any of those fields, or remove the assistant
- upload-file, list-files and delete-file — the files assistants search over
Requirements
An OpenAI API key, supplied as OPENAI_API_KEY or passed to the server as --openai-api-key. Install with pip install . from the repository; the client then runs mcp-server-openai over stdio. Model usage is billed by OpenAI at their published rates.
Setup effort
Build from source — git clone https://github.com/snilld-ai/openai-assistant-mcp
USE CASES
01
Standing up an assistant over a document set
Upload the files, then create the assistant with those file IDs and file_search enabled.
02
Adjusting an assistant that answers badly
Read its current instructions and model, then update them in place.
03
Getting a second opinion mid-conversation
Put the same question to a specific OpenAI model and compare the answer.
04
Cleaning up an account
List the assistants and files, then delete the ones nothing uses.
9 tools
Listed in project docs. Your client may show a different list.
list-assistants
List all available assistants.
retrieve-assistant
Retrieve an assistant by its ID.
create-assistant
Create a new assistant.
update-assistant
Update an existing assistant.
delete-assistant
Delete an assistant by its ID.
upload-file
Upload a file for use with assistants.
list-files
List all files available for assistants.
delete-file
Delete a file by its ID.
BEFORE YOU INSTALL
What to know first
Tools you get
9 tools named in the project's documentation.
Requires
- pip on PATH
- OPENAI_API_KEY in your environment
Where it runs
The documented client runs the server on your machine.
INSTALLATION
Connect your client
Server setup
TERMINAL
git clone https://github.com/snilld-ai/openai-assistant-mcp
cd openai-assistant-mcp
pip install .MCP Clients
Cursor
CURSOR · JSON
{
"mcpServers": {
"openai-server": {
"command": "mcp-server-openai",
"args": [],
"env": {
"PYTHONPATH": "<YOUR_PYTHONPATH>",
"OPENAI_API_KEY": "<YOUR_OPENAI_API_KEY>"
}
}
}
}<YOUR_PYTHONPATH> is the absolute path of your clone (it goes on PYTHONPATH), and <YOUR_OPENAI_API_KEY> is your OpenAI API key. mcp-server-openai is the command the pip install step puts on your PATH.
