Skip to main content
Glama
kinfey

budget-evaluation-mcp

by kinfey
OverviewSchemaRelated ServersScoreDiscussions
Python
Hybrid

Engineering Budget Evaluation Agent

Microsoft Agent Framework + GitHub Copilot SDK (gpt-5.6-sol) solution for evaluating engineering budget workbooks through MCP and Microsoft Teams.

The calculation layer is deterministic. It reads PER Budget, QS Estimation, Vendor Quotation, MAPPING RULES, Historical Unit Rates, and Summary; maps every priced vendor WBS row; and normalizes category and total rates to USD/Sqm. GPT-5.6 Sol produces the executive narrative and answers follow-up questions from the structured results.

Architecture

┌───────────────────────────────┐
│  User / Microsoft Teams      │
│  Uploads engineering Excel   │
│  workbook for budget review   │
└──────────────┬────────────────┘
               │
               ▼
┌───────────────────────────────┐
│  Teams App                    │
│  - Receives file attachment   │
│  - Downloads workbook         │
│  - Stores conversation report │
│  - Sends follow-up questions  │
└──────────────┬────────────────┘
               │
               ▼
┌───────────────────────────────┐
│  MCP Client                   │
│  - Calls evaluate_budget_     │
│    workbook tool              │
│  - Calls ask_budget_report    │
│  - Uses Streamable HTTP      │
└──────────────┬────────────────┘
               │
               ▼
┌───────────────────────────────┐
│  Budget Evaluation MCP        │
│  - /mcp endpoint             │
│  - Validates workbook         │
│  - Caches reports            │
│  - Returns structured data   │
└──────────────┬────────────────┘
               │
               ▼
┌───────────────────────────────┐
│  Deterministic Evaluation     │
│  - Reads PER / QS / History  │
│  - Maps vendor WBS rows      │
│  - Normalizes USD/Sqm        │
│  - Calculates variance       │
│  - Produces vendor decisions  │
└──────────────┬────────────────┘
               │
               ▼
┌───────────────────────────────┐
│  Structured Report Model     │
│  - Project metadata          │
│  - Vendor totals              │
│  - Category comparisons       │
│  - Warnings and unmatched     │
│    items                     │
└──────────────┬────────────────┘
               │
               ▼
┌───────────────────────────────┐
│  AI Summary Layer             │
│  - GitHub Copilot model      │
│  - Generates executive       │
│    recommendation            │
│  - Answers follow-up Q&A     │
└──────────────┬────────────────┘
               │
               ▼
┌───────────────────────────────┐
│  Response Back to User       │
│  - Evaluation card           │
│  - Summary narrative         │
│  - Chat follow-up answers    │
└───────────────────────────────┘

Key components

  • teams_app/ — Microsoft Teams bot and UI integration

  • budget_agent/server.py — MCP server exposing evaluation and report-query tools

  • budget_agent/evaluator.py — Excel validation and deterministic cost comparison logic

  • budget_agent/agent.py — GitHub Copilot-based narrative and Q&A generation

  • tests/ — validation and regression tests for evaluator behavior

Related MCP server: Tri-Tender Pricing MCP

Evaluation rules

  • Vendor total versus PER Budget:

    • |variance| <= 10%: approve

    • 10% < |variance| <= 20%: conditional approval

    • |variance| > 20%: reject

  • Vendor mapped category versus QS Estimate:

    • |variance| <= 5%: reasonable

    • 5% < |variance| <= 15%: review required

    • |variance| > 15%: significant concern

  • Vendor mapped category unit rate versus historical USD/Sqm: same ±5% / ±15% thresholds.

Unmapped priced rows and missing benchmarks are reported explicitly.

OUTPUT-style dashboard

The MCP report and Teams Adaptive Card follow the workbook OUTPUT tab:

  • Page 1 — Vendor total versus PER Budget

  • Page 2 — Vendor mapped category amount versus QS Estimate

  • Page 3 — Vendor mapped category USD/Sqm versus Historical Unit Rate

Traffic-light indicators are consistent across the structured JSON and Teams UI:

  • 🟢 Green — approve / reasonable

  • 🟡 Yellow — conditional approval / review required

  • 🔴 Red — reject / significant cost concern

  • ⚪ Gray — benchmark missing

Local validation

conda activate agentdev
pip install -e '.[dev,azure]'
pytest -q
ruff check budget_agent tests scripts
python -m budget_agent.server

The MCP endpoint is http://localhost:8000/mcp; health is http://localhost:8000/healthz.

Teams app

cd teams_app
npm install
npm run build

Set the values in teams_app/.env.example, then run npm start. Upload an .xlsx file in Teams; the bot calls evaluate_budget_workbook over MCP and keeps the returned report ID for follow-up questions in the same conversation.

Azure deployment

conda activate agentdev
export AZURE_SUBSCRIPTION_ID="$(az account show --query id -o tsv)"
export AZURE_RESOURCE_GROUP="rg-budget-agent"
export AZURE_LOCATION="eastus2"
export ACA_SANDBOX_GROUP="aca-sbx-budget-agent"
export COPILOT_GITHUB_TOKEN="<GitHub token authorized for Copilot requests>"
python scripts/deploy_sandbox.py

export MCP_URL="$(python -c 'import json; print(json.load(open(".azure/sandbox-deployment.json"))["mcp_url"])')"
bash scripts/deploy_teams.sh

The scripts create an Azure Container Apps Sandbox MCP service and a single-tenant Teams bot, enable the Microsoft Teams channel, and generate .azure/engineering-budget-teams-app.zip for Teams upload. The .azure/ directory and all deployment-specific identifiers are intentionally excluded from source control.

This server cannot be installed

F
license - not found
-
quality - not tested
C
maintenance

How are these scores calculated?

Maintenance

–Maintainers
–Response time
–Release cycle
–Releases (12mo)
Commit activity

Resources

Unclaimed servers have limited discoverability.

Looking for Admin?

If you are the server author, to access and configure the admin panel.

Related MCP Servers

  • ClaudeHopper

    Arborist-ai
    A
    license
    -
    quality
    D
    maintenance
    AI-powered MCP server that enables Claude and other LLMs to interact directly with construction documents, drawings, and specifications through advanced RAG and hybrid search capabilities.
    7
    MIT

  • Tri-Tender Pricing MCP

    sendsta
    F
    license
    -
    quality
    D
    maintenance
    An MCP server designed to automate tender and RFQ pricing by extracting requirements from documents and building structured pricing models. It enables users to calculate final costs, compare market rates, and generate styled HTML pricing reports for PDF export.

  • actual-mcp-server

    agigante80
    A
    license
    -
    quality
    A
    maintenance
    MCP server that connects AI assistants to Actual Budget for budget management, enabling natural language queries, transaction creation, and spending analysis.
    1,497
    39
    MIT

View all related MCP servers

Related MCP Connectors

  • PlanExe

    MCP server for generating rough-draft project plans from natural-language prompts.

  • TokenOracle

    Hosted MCP server for LLM cost estimation, model comparison, and budget-aware routing.

  • futureagievals-mcp

    A paid remote MCP for AI SDK eval dashboard, built to return verdicts, receipts, usage logs, and aud

View all MCP Connectors

Latest Blog Posts

MCP directory API

We provide all the information about MCP servers via our MCP API.

curl -X GET 'https://glama.ai/api/mcp/v1/servers/kinfey/hack_demo'

If you have feedback or need assistance with the MCP directory API, please join our Discord server