prompt_for_user_inputLABSCO SUMMARY
It changes when the model stops.
The value is not the tool but the instruction that goes with it: told never to end a turn without asking, the assistant surfaces its uncertainty as a question instead of guessing and finishing. Copy the suggested system prompt when you install it, and keep the questions in prose — Markdown will not render where the prompt appears.
What it is
An MCP server that turns the conversation around: instead of the assistant finishing and waiting, it can call a tool that puts a question in front of you in the editor and blocks on your reply.
What you get
- A question surfaced to the user, with the answer returned to the model —
prompt_for_user_input - A suggested system prompt in the README that makes the behaviour the default: keep asking through the tool rather than ending the turn
- Guidance that the prompts stay plain prose, since Markdown will not render in the input surface
- An HTTP server you start yourself —
uv run uvicorn mcp_server:app --port 4444
Requirements
Python with uv, running the server through uvicorn on port 4444. One shared secret is needed on both sides: PROMPT_FOR_USER_INPUT_MCP_AUTH_TOKEN must be set for the server and in the Cursor MCP configuration that talks to it. The project is prompt-for-user-input-mcp 1.0.0.
Setup effort
Build from source — git clone https://github.com/goldensansapple/prompt-for-user-input-mcp.git
USE CASES
01
Getting asked instead of guessed at
Have the assistant put its open question to you before it commits to an approach.
02
Keeping a long task moving
Let the model check in when it gets stuck rather than ending the turn with a half-finished answer.
03
Confirming before something irreversible
Require an explicit answer in the editor before the assistant proceeds.
1 tool
Listed in project docs. Your client may show a different list.
BEFORE YOU INSTALL
What to know first
Tools you get
1 tool named in the project's documentation.
Requires
PROMPT_FOR_USER_INPUT_MCP_AUTH_TOKEN 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/goldensansapple/prompt-for-user-input-mcp.git
cd prompt-for-user-input-mcp
uv run uvicorn mcp_server:app --port 4444MCP Clients
Cursor
CURSOR · JSON
{
"mcpServers": {
"prompt-for-user-input-mcp": {
"url": "http://127.0.0.1:4444/mcp",
"headers": {
"Authorization": "Bearer ${env:PROMPT_FOR_USER_INPUT_MCP_AUTH_TOKEN}"
}
}
}
}This is the repository's own cursor-mcp-config.json; the server has to be running on port 4444 first.
The README says both the server and the Cursor configuration need PROMPT_FOR_USER_INPUT_MCP_AUTH_TOKEN in their environment, set to the same token — Cursor sends it as Bearer ${env:PROMPT_FOR_USER_INPUT_MCP_AUTH_TOKEN}. The repository's generate_token.py prints a random token you can use.
