TriTender Pricing

This Tri‑Tender Pricing MCP server provides a production‑ready, locally deterministic pricing engine for tenders and RFQs. It detects pricing requirements, builds a pricing model from tender rules and company rates, computes final prices with markups and VAT, and can generate a styled HTML report ready for PDF export or market comparisons. It is designed to run inside MCP hosts and integrates with Tri‑Tender workflows for preview and export.

How to use

You run this MCP server alongside your MCP client to perform end‑to‑end pricing workflows. Start the server locally or remotely, then register it with your Tri‑Tender environment. From the MCP client you can trigger the pricing workflow end to end: extract pricing requirements from tender packs, build a structured pricing model, calculate line totals and totals including VAT, optionally compare against market prices, and render a styled HTML report for preview or PDF export. Use the client to request pricing data when you have a tender or pricing schedule ready for analysis.

When you request a pricing run, you’ll work with tools exposed by the server, such as extracting requirements, building models, calculating prices, fetching market prices, and rendering a report. The output is designed to be easily consumed by Tri‑Tender previews, with data structured for LLM workflows and a ready‑to‑export HTML report.

How to install

Prerequisites you need on your host: a Python 3.x runtime and access to install Python packages via pip.

# Create and activate a virtual environment (recommended)
python -m venv .venv
source .venv/bin/activate  # on Unix or macOS
# Windows "venv\Scripts\activate" in PowerShell or CMD

# Install dependencies
pip install -r requirements.txt

# Verify the runtime environment
fastmcp version

Running the MCP server

Start the server using the FastMCP runner. By default, the server uses stdio transport, which works well for local hosts and developer machines. If you need SSE/HTTP transport, you can run with the appropriate options.

fastmcp run server.py

If you prefer to run the Python script directly, you can start it like this:

python server.py

If your host supports SSE/HTTP, you can expose the server on a network interface and port. The following command demonstrates enabling SSE transport and binding to all interfaces on port 8000:

fastmcp run --transport sse --host 0.0.0.0 --port 8000 server.py

Configuration and registration with Tri‑Tender client

Register the MCP server with your Tri‑Tender client using a local, standard MCP configuration. You can choose either a local stdio setup or a remote HTTP/SSE setup.

{
  "mcpServers": {
    "tri_tender_pricing_mcp": {
      "command": "python",
      "args": ["server.py"],
      "env": {}
    }
  }
}

Remote HTTP/SSE registration example

If you deploy the server remotely and expose it via SSE/HTTP, use the network URL to connect from the Tri‑Tender client.

{
  "mcpServers": {
    "tri_tender_pricing_mcp": {
      "url": "http://YOUR-HOST:8000/sse"
    }
  }
}

Available tools

detect_pricing_requirements

Reads input documents (PDF, DOCX, XLSX), cleans text, classifies document type, extracts pricing sections, and returns a structured summary including currency and a rough pricing items skeleton.

build_model

Converts free‑text descriptions and tender/company rules into a structured pricing model with items, base rates, quantities, and meta information like currency and VAT.

calculate

Applies line‑level calculations to produce totals, VAT amounts, and grand totals; returns detailed item results and overall totals.

market_prices

Fetches or simulates market prices for items to enable comparison against internal pricing.

render_report

Renders a styled HTML report from pricing data, ready for export or preview.

final_output

Wraps the rendered HTML into a final structure for Tri‑Tender desktop or Dyad templates.