survey-collector-mcp 0.1.0

MCP server for survey collection with LLM-powered intelligent follow-up questions

Survey Collector MCP Server

MCP server for survey collection with LLM-powered intelligent follow-up questions.

Features

• 📋 Survey Collection: Send questions to groups and collect replies
• 🤖 LLM-powered Follow-up: Automatically generate follow-up questions based on replies
• 📊 Criteria Evaluation: Multiple evaluation criteria (all replied, majority agree, etc.)
• 📝 Summary Reports: Auto-generate Markdown summary reports

Installation

Via pip

pip install survey-collector-mcp

Via uvx (recommended)

uvx survey-collector-mcp

Usage

Local Mode (stdio)

Add to your mcp.json:

{
  "mcpServers": {
    "survey-collector": {
      "command": "uvx",
      "args": ["survey-collector-mcp"]
    }
  }
}

Remote Mode (SSE)

Start the server:

survey-collector --transport sse --host 0.0.0.0 --port 8001

Client configuration:

{
  "mcpServers": {
    "survey-collector": {
      "url": "http://localhost:8001/sse"
    }
  }
}

Tools

collect_survey_info

Execute a complete survey workflow with intelligent follow-up.

Parameters:

Parameter	Type	Required	Description
chat_id	string	Yes	Group chat ID
members	string	Yes	Member names, comma-separated
question	string	Yes	Survey question
criteria	string	No	Evaluation criteria (default: all_replied)
follow_up_rule	string	No	Follow-up rule in natural language
max_rounds	int	No	Max follow-up rounds (default: 5)
timeout_per_round	int	No	Timeout per round in seconds (default: 600)

Example:

Question: "周一上午有没有时间打球？"
Follow-up rule: "如果对方说没时间，每次追问日期往后推一天"

Result:
- Round 1: "周一上午有没有时间打球？"
- Round 2: "周二上午有没有时间打球？" (LLM generated)
- Round 3: "周三上午有没有时间打球？"

Configuration

Set environment variables or create .env file:

# HITL Platform
HITL_SERVICE_URL=https://hitl.woa.com/api

# LLM API (for intelligent follow-up)
LLM_API_BASE=https://api.deepseek.com/v1
LLM_API_KEY=your-api-key
LLM_MODEL=deepseek-chat

# Debug mode
DEBUG=false

License

MIT License