Sales Proposal MCP Server

An MCP (Model Context Protocol) server that integrates with Claude Desktop to analyze company call transcripts and generate comprehensive sales proposals with MEDDIC evaluations.

Installation

Install from Source (Recommended)

Clone the repository and install the MCP server globally:

git clone https://github.com/MaPa07/sales-proposal-mcp.git
cd sales-proposal-mcp
npm install -g .

NPM Package (Coming Soon)

Once published to npm, you can install directly:

npm install -g @badlieutenant/sales-proposal-mcp

Claude Desktop Configuration

After installation, add the server to your Claude Desktop configuration:

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "sales-proposal": {
      "command": "sales-proposal-mcp",
      "env": {
        "SALES_PROPOSAL_ACCESS_TOKEN": "your-access-token-here"
      }
    }
  }
}

Important: Replace "your-access-token-here" with the actual access token shared via 1Password. The server will not function without a valid access token.

Claude Project Setup (Recommended)

After installing the MCP server, create a new Claude project with the following custom instructions for optimal proposal generation:

This Project is designed to generate customer proposals. Each chat will be for a different company. 

Whenever someone requests a proposal to be created for a company (domain), don't go out and create a generic proposal. 

Instead, always ping the proposal MCP with the company domain. This will take 60-120 seconds and you will get a response.

The response will have 2 parts.

- the actual proposal and discovery
The key here is the Current state and Desired state sections in bulleted lists that have as much details / metrics as we find in the initial transcripts
We don't need timelines or a generic summary of what the company does. Keep in mind that the proposal should ultimately be shared with the customer and doesn't need to contain information that's already obvious to the customer such as what industry they are in.
When providing key stakeholders make sure to clearly separate stakeholders identified at the company vs. at MadKudu


- a rating for the proposal and a rating including feedback for the seller.
Make sure this stays actionable. Concise lists of gaps and action items should stand out.


When you get the response, create 2 artifacts, one for each section.
Make sure to keep the summaries concise, metric focused and not too verbose. Bulleted lists are a great format. 


The user can ask follow up questions and ask about modifications.

Usage

Once configured, restart Claude Desktop and you can ask Claude to analyze company calls:

• "Analyze Stripe's call data and create a comprehensive sales proposal"
• "Generate a sales proposal for abtasty.com focusing on enterprise features"
• "Analyze monday.com and provide MEDDIC evaluation"

The analysis typically takes 60-120 seconds to complete.

Features

• Business Analysis: Comprehensive company analysis from call transcripts
• Sales Proposals: AI-generated proposals tailored to company needs
• MEDDIC Evaluation: Structured sales qualification framework
• Additional Insights: Strategic recommendations and next steps

Tool Schema

The server exposes one tool:

{
  name: 'analyze_company_calls',
  description: 'Analyze company call transcripts and generate comprehensive sales proposal with MEDDIC evaluation',
  inputSchema: {
    type: 'object',
    properties: {
      domain: {
        type: 'string',
        description: 'Company domain like abtasty.com (required)'
      },
      additionalContext: {
        type: 'string', 
        description: 'Optional context like "focus on enterprise features"'
      }
    },
    required: ['domain']
  }
}

Configuration

Access Token (Required)

This MCP server requires an access token to control usage and costs. The token should be:

• Shared via 1Password with your team
• Added to your Claude Desktop configuration as shown above
• Kept secure and not committed to version control

Requirements

• Node.js 18 or higher
• Valid access token for the service

Error Handling

The server gracefully handles:

• Domain not found errors
• Service timeouts
• Network connectivity issues
• Invalid domain formats

All errors include descriptive messages to help users understand what went wrong.

License

MIT