aws-nasmatchstockprofile-mcp

aburmd/aws-nasmatchstockprofile-mcp

3.1

If you are the rightful owner of aws-nasmatchstockprofile-mcp and would like to certify it and/or have it hosted online, please leave a comment on the right or send an email to dayong@mcphub.com.

The Model Context Protocol (MCP) server is designed to facilitate seamless integration and automation of workflows by leveraging AI agents to manage infrastructure and runtime code.

Tools
2
Resources
0
Prompts
0

Welcome to your CDK TypeScript project

This is a blank project for CDK development with TypeScript.

The cdk.json file tells the CDK Toolkit how to execute your app.

Useful commands

  • npm run build compile typescript to js
  • npm run watch watch for changes and compile
  • npm run test perform the jest unit tests
  • npx cdk deploy deploy this stack to your default AWS account/region
  • npx cdk diff compare deployed stack with current state
  • npx cdk synth emits the synthesized CloudFormation template

aws-nasmatchstockprofile-mcp/ │── cdk/ # AWS CDK Infrastructure (TypeScript) │ ├── bin/ │ │ └── aws-nasmatchstockprofile-mcp.ts # CDK entry point │ ├── lib/ │ │ └── aws-nasmatchstockprofile-mcp-stack.ts # Main stack │ ├── package.json │ ├── cdk.json │ └── tsconfig.json │ │── src/ # Python Runtime (MCP + AI Agent) │ ├── mcp_server/ │ │ ├── init.py │ │ ├── main.py # MCP server entry │ │ ├── excel_handler.py # Logic for Excel updates │ │ ├── data_extractor.py # Logic for parsing website Excel │ │ └── config.py # Configurations (e.g. S3 bucket, env vars) │ │ │ ├── agent/ │ │ ├── init.py │ │ ├── orchestrator.py # AI agent orchestration │ │ └── prompts.py # System prompts, reasoning templates │ │ │ ├── tests/ │ │ └── test_excel_handler.py │ │ │ └── requirements.txt │ │── .gitignore │── README.md

git remote set-url origin https://aburmd@github.com/aburmd/aws-nasmatchstockprofile-mcp.git

ssh-keygen -t ed25519 -C "aburmd@gmail.com"

cat ~/.ssh/id_ed25519.pub

Copy the output → Go to GitHub → Settings → SSH and GPG keys → New SSH key → Paste.

git remote remove origin git remote add origin git@github.com:aburmd/aws-nasmatchstockprofile-mcp.git

ssh -T git@github.com

You should see: "Hi aburmd! You've successfully authenticated."

git push -u origin main

Here’s a single, copy-pasteable master prompt you can give to your code assistant to scaffold the entire repo. It bakes in every requirement: CDK in TypeScript, all runtime in Python, API Gateway WebSocket for MCP-style tools, and Step Functions + ECS Fargate for ETL.

🔧 MASTER PROMPT (copy–paste below this line)

You are an expert AWS/CDK solution engineer and senior Python backend developer. Generate a production-ready Git repository that implements the following, with all Infrastructure as Code in TypeScript (CDK v2) and all runtime code in Python 3.11.

High-Level Goal

  • Build an MCP-style tool endpoint over API Gateway WebSocket so any AI agent can call tools like:

    • mapping.propose (Bedrock → Claude; column mapping with confidences)
    • mapping.apply (DynamoDB upserts)
    • excel.replace_table (S3 + openpyxl; writes a table into a target sheet of an Excel template)

  • Build a daily deterministic ETL that:

    • Reads raw/source.xlsx from S3
    • Normalizes to CSV/Parquet (staging/)
    • Publishes a curated master.xlsx to curated/ using an Excel template

  • Orchestration via Step Functions running an ECS Fargate task (Python container) for heavy ETL steps.

  • All configs/versioned artifacts live in S3 (config/, raw/, staging/, curated/).

Repository Layout (create these paths & files)

aws-excel-mcp/
  README.md
  .gitignore
  package.json
  cdk.json
  tsconfig.json
  bin/app.ts
  lib/
    kms-stack.ts
    storage-stack.ts
    ecr-stack.ts
    network-stack.ts
    ecs-runner-stack.ts
    ws-mcp-stack.ts
    stepfn-stack.ts
    events-stack.ts
  containers/
    excel-runner/
      Dockerfile
      app/handler.py
  runtime/
    mcp_router/handler.py
    mapping_propose/handler.py
    mapping_apply/handler.py
    excel_replace/
      handler.py
      requirements.txt
  s3_bootstrap/config/
    mapping.yaml
    master_template.xlsx   # put a placeholder binary or describe in README to upload
  .github/workflows/cdk-deploy.yml

Core Requirements

Infrastructure (CDK TypeScript)

  • KMS CMK (alias/excel-pipeline-kms) for S3 + DynamoDB encryption.

  • S3 bucket (versioned, KMS-encrypted) with prefixes: raw/, staging/, curated/, config/.

  • DynamoDB:

    • MappingOverrides (PK dataset_id STRING, SK source_col STRING).
    • WsConnections (PK connection_id STRING) for WebSocket connection management.

  • API Gateway WebSocket:

    • Stage prod, routes: $connect, $disconnect, $default, plus explicit routes mapping.propose, mapping.apply, excel.replace_table.
    • Router Lambda (Python) handles connect/disconnect/default and dispatches to tool Lambdas; responds over @connections API.

  • Tool Lambdas (Python):

    • mapping_propose: calls Amazon Bedrock (Claude 3.5 Sonnet model id anthropic.claude-3-5-sonnet-20240620-v1:0). Input: source_cols[], canonical_cols[]. Output: JSON array [{from,to,confidence}]. Return strict JSON only.
    • mapping_apply: upserts items into MappingOverrides.
    • excel_replace: loads config/master_template.xlsx from S3, writes tabular data to given sheet/start_cell with openpyxl, and saves to curated/master.xlsx.

  • VPC with public + private subnets; VPC endpoints for S3 (Gateway) and Bedrock Runtime (Interface).

  • ECR repository for the ETL container: excel-runner.

  • ECS Fargate:

    • Cluster + TaskDefinition (1 vCPU/2GB) to run containers/excel-runner.

  • Step Functions state machine:

    1. Transform → run ECS task with args to read raw/source.xlsx, normalize to staging/orders.parquet + staging/orders.csv.
    2. ProposeMapping → call mapping_propose Lambda.
    3. ApplyMapping → call mapping_apply Lambda.
    4. Publish → run ECS task to create/overwrite staging/master_draft.xlsx and promote to curated/master.xlsx via template.

  • EventBridge schedule: daily 06:05 AM PT (cron hour 13 UTC), triggers the state machine.

  • IAM least privilege:

    • Router: execute-api:ManageConnections, lambda:InvokeFunction for tool Lambdas, RW on WsConnections.
    • Tool Lambdas: scoped S3, DDB; Bedrock invoke for mapping_propose.
    • ECS task role: scoped S3 RW.

  • Use @aws-cdk/aws-apigatewayv2-alpha and @aws-cdk/aws-apigatewayv2-integrations-alpha for WebSocket + Lambda integration.

  • Use @aws-cdk/aws-lambda-python-alpha PythonFunction to bundle Python Lambdas and requirements.txt for excel_replace.

Runtime Code (Python 3.11)

  • Router Lambda:

    • On $connect: store connection_id in DDB, send handshake JSON { "mcp": "hello", "tools": ["mapping.propose","mapping.apply","excel.replace_table"] }.
    • On $disconnect: delete connection_id.
    • On $default or explicit tool route: parse body JSON; figure out tool name (either routeKey or body.tool); invoke the right tool Lambda synchronously; respond to client with { ok: true, tool, result }.

  • mapping_propose Lambda:

    • Prompt template to strictly return JSON array; parse LLM output robustly; if parse fails, return [].

  • mapping_apply Lambda:

    • Input: { "dataset_id": "orders_v1", "mapping": [{from,to,confidence}] }. Upsert to DDB; return count.

  • excel_replace Lambda:

    • Input: {template_key, out_key, sheet, start_cell, data:[{...}]}.
    • Use openpyxl to clear a reasonable range, write headers + rows, save to S3.

  • ECS runner container (excel-runner):

    • CLI with --op transform and --op publish.
    • transform: read Excel from raw/, apply mapping.yaml rules (rename/dtypes/filters/dedupe), write Parquet + CSV to staging/.
    • publish: read CSV + template from S3, write to target sheet/cell, put draft to staging/master_draft.xlsx and final to curated/master.xlsx.

Config Files

  • s3_bootstrap/config/mapping.yaml (sample):

    dataset_id: "orders_v1"
source:
  sheet: "Export_Orders"
  sheets:
    - name: "Export_Orders"
      rename:
        "Order Id": order_id
        "Customer Name": customer
        "Order Date": order_date
        "Total": total_amount
      dtypes:
        order_date: datetime
        total_amount: float
      dedupe_keys: [order_id]
targets:
  - sheet: "Raw_Orders"
    mode: replace_table
ai:
  min_confidence_auto_apply: 0.85

  • Put a placeholder master_template.xlsx (or document that the user should upload theirs).

Dockerfile (Python)

  • containers/excel-runner/Dockerfile: based on python:3.11-slim; install pandas, openpyxl, pyarrow, boto3, pyyaml.

GitHub Actions (optional)

  • cdk-deploy.yml that builds CDK, synthesizes, and deploys (OIDC role assumed via secrets).

Non-Functional Requirements

  • Clear, defensive error handling (no crashes on malformed JSON or empty datasets).
  • Logging with useful context: request IDs, S3 keys, row counts, confidence stats.
  • Idempotent publishes (only promote to curated/ after successful run).
  • Versioning: rely on S3 versioning for rollbacks.
  • Security: least-privilege IAM, KMS at rest, private VPC traffic to Bedrock/S3 where possible.

README.md (include)

  • High-level architecture diagram (ASCII ok).

  • Setup steps:

    1. npm install, cdk bootstrap
    2. cdk deploy Kms Storage Ecr Net
    3. Build & push excel-runner image to ECR
    4. cdk deploy EcsRunner WsMcp StepFn Events
    5. Upload s3_bootstrap/config/* and a sample raw/source.xlsx

  • How to test WebSocket with wscat:

    wscat -c wss://<api-id>.execute-api.<region>.amazonaws.com/prod
> {"tool":"mapping.propose","source_cols":["Order Id","Cust Name"],"canonical_cols":["order_id","customer"]}

  • How to start a manual Step Functions execution and where outputs land in S3.

  • Cost notes (API GW WS pay-per-message; no ALB).

Acceptance Criteria

  1. cdk synth and cdk deploy succeed with no manual edits beyond region/account.
  2. WebSocket connection succeeds; $connect returns tool list; calling mapping.propose returns a JSON array (mockable if Bedrock isn’t enabled).
  3. Running the state machine processes a sample raw/source.xlsx into staging/*.parquet/.csv and publishes curated/master.xlsx.
  4. excel.replace_table writes to a hidden or staging sheet without corrupting formulas in the template (document that the visible sheets reference Raw_Orders).
  5. IAM policies are restricted to necessary resources only.

Coding Standards

  • Python: 3.11, type hints where reasonable, small functions, explicit returns, PEP8.
  • CDK TS: clear stack boundaries, export key ARNs/URLs via stack class properties, no hardcoded ARNs.
  • Use environment variables for Bucket name, DDB table, Model ID.
  • Tool Lambdas return strict JSON; router always wraps in {ok, tool, result}.

Bonus (if time allows)

  • Add a sfn.start tool (Lambda) so agents can kick the batch via WebSocket.
  • Step Functions choice state to require human approval if any mapping confidence < 0.85.
  • API GW authorizer (Cognito or IAM) for WebSocket.

Deliverables: fully generated repo tree with all files populated per above, ready to commit and deploy. Do not omit any file listed. Print the final repository tree and any post-deploy instructions at the end.

🔧 END MASTER PROMPT

Want me to tailor the sample mapping.yaml to your exact column names? If you share a dummy source.xlsx header row and target sheet name, I’ll plug them in.

mkdir aws-nasmatchstockprofile-mcp cd aws-nasmatchstockprofile-mcp

Initialize CDK with TypeScript

cdk init app --language typescript

install cdk

node -v npm -v

If not installed, grab it via Node.js official site or Homebrew: brew install node

npm install -g aws-cdk

cdk bootstrap aws://654654547262/us-east-1 cdk deploy

npm run build

Deploy core data plane first

cdk deploy BaseInfra

Deploy networking + ECS runner

cdk deploy Net EcsRunner

Deploy WebSocket + Lambdas

cdk deploy WsMcp

Deploy Step Functions + schedule

cdk deploy StepFn Events

Got it 👍 — let’s walk through step by step so you can learn how to approach this with MCP (Model Context Protocol) and AWS CDK.

We’ll treat this as:

  1. Business Requirement → Update Excel based on another Excel.
  2. Architecture → AWS cost-effective, scalable, Git-based infra.
  3. MCP Role → Automate infra + runtime workflows via AI agent.
  4. Step-by-Step CDK Build → Start small (base stack), then extend.

1. Business Requirement

  • You have Excel File A (source) and Excel File B (target).
  • The system should update B based on values from A (rules defined by you).
  • This must be automated, repeatable, and cloud-based.

2. Architecture (AWS Components)

  • S3 → Store input & output Excel files.
  • Lambda (Python runtime) → Process Excel updates (using openpyxl or pandas).
  • EventBridge / S3 Trigger → Kick off processing when new file uploaded.
  • IAM Roles → Secure access.
  • (Optional) Step Functions if workflow grows complex.

This is all CDK-driven, deployed via Git.

3. Role of MCP (Model Context Protocol)

  • MCP allows you to define actions for AI agents (like me) to orchestrate infra and runtime code.

  • Example:

    • Define a tool in MCP for "Deploy CDK stack".
    • Define another for "Process Excel".

  • Then your AI agent can decide autonomously: upload file → trigger Lambda → update Excel → store output → notify you.

So MCP is the glue to:

  • Describe capabilities (e.g., "Excel Processor").
  • Expose them to AI agent (so it knows it can invoke CDK/Lambda).
  • Automate workflows (instead of you clicking manually).

4. Step-by-Step CDK Build Plan

Here’s how we’ll build incrementally:

Step 1: Base Stack

  • VPC (if needed), S3 bucket(s).
  • IAM roles & policies.
  • Output: "We can upload Excel files."

Step 2: Lambda Runtime

  • Python code (using openpyxl) packaged & deployed.
  • Connected to S3.
  • Output: "Lambda runs when file uploaded."

Step 3: Event Trigger

  • Configure S3 → EventBridge → Lambda.
  • Output: "Automation kicks off on upload."

Step 4: Excel Processing Logic

  • Lambda reads fileA.xlsx & fileB.xlsx, applies rules, writes back to S3.
  • Output: "Excel files updated automatically."

Step 5: MCP Exposure

  • Define MCP tool(s) → “Upload Excel”, “Run Processor”, “Deploy Stack”.
  • Output: "AI agent can handle Excel updates end-to-end."

💡 Cost:

  • S3 = pennies per GB.
  • Lambda = free tier + $0.20 per 1M requests.
  • EventBridge = $1 per 1M events.
  • IAM roles = free. 👉 Essentially close to $0 monthly unless you process huge Excel files.

Do you want me to start with Step 1 (Base Stack CDK code) and explain the CDK resources we’ll add before we write the code?

aws-nasmatchstockprofile-mcp

Foundational CDK app for an MCP-controlled, cost-effective Excel processing pipeline on AWS.

Step 1 (this repo state)

  • KMS CMK (alias: alias/excel-pipeline-kms)
  • S3 bucket (versioned, KMS-encrypted)
  • DynamoDB tables:
    • MappingOverrides (PK: dataset_id, SK: source_col)
    • WsConnections (PK: connection_id)

Install & Deploy

# from repo root
npm install
npm run build
npm run synth
npm run deploy -- BaseInfra


# after step4
aws-nasmatchstockprofile-mcp/
  bin/app.ts
  lib/base-infra.ts              # add WebSocket API + permissions here
  lambda/
    processor/                   # (already from step2/3)
      handler.py
      requirements.txt
    wsmcp/
      on_connect.py
      on_disconnect.py
      on_message.py
      requirements.txt



Got it 👍 — let’s walk through step by step so you can learn how to approach this with MCP (Model Context Protocol) and AWS CDK.

We’ll treat this as:

Business Requirement → Update Excel based on another Excel.

Architecture → AWS cost-effective, scalable, Git-based infra.

MCP Role → Automate infra + runtime workflows via AI agent.

Step-by-Step CDK Build → Start small (base stack), then extend.

1. Business Requirement

You have Excel File A (source) and Excel File B (target).

The system should update B based on values from A (rules defined by you).

This must be automated, repeatable, and cloud-based.

2. Architecture (AWS Components)

S3 → Store input & output Excel files.

Lambda (Python runtime) → Process Excel updates (using openpyxl or pandas).

EventBridge / S3 Trigger → Kick off processing when new file uploaded.

IAM Roles → Secure access.

(Optional) Step Functions if workflow grows complex.

This is all CDK-driven, deployed via Git.

3. Role of MCP (Model Context Protocol)

MCP allows you to define actions for AI agents (like me) to orchestrate infra and runtime code.

Example:

Define a tool in MCP for "Deploy CDK stack".

Define another for "Process Excel".

Then your AI agent can decide autonomously: upload file → trigger Lambda → update Excel → store output → notify you.

So MCP is the glue to:

Describe capabilities (e.g., "Excel Processor").

Expose them to AI agent (so it knows it can invoke CDK/Lambda).

Automate workflows (instead of you clicking manually).

4. Step-by-Step CDK Build Plan

Here’s how we’ll build incrementally:

Step 1: Base Stack

VPC (if needed), S3 bucket(s).

IAM roles & policies.

Output: "We can upload Excel files."

Step 2: Lambda Runtime

Python code (using openpyxl) packaged & deployed.

Connected to S3.

Output: "Lambda runs when file uploaded."

Step 3: Event Trigger

Configure S3 → EventBridge → Lambda.

Output: "Automation kicks off on upload."

Step 4: Excel Processing Logic

Lambda reads fileA.xlsx & fileB.xlsx, applies rules, writes back to S3.

Output: "Excel files updated automatically."

Step 5: MCP Exposure

Define MCP tool(s) → “Upload Excel”, “Run Processor”, “Deploy Stack”.

Output: "AI agent can handle Excel updates end-to-end."

💡 Cost:

S3 = pennies per GB.

Lambda = free tier + $0.20 per 1M requests.

EventBridge = $1 per 1M events.

IAM roles = free.
👉 Essentially close to $0 monthly unless you process huge Excel files.






aws lambda invoke --function-name BaseInfra-ExcelProcessorFnB7B43EA3-fRqa6Yu3FXTU \
--payload '{
    "source_key":"source/Portfolio_Positions_Aug-21-2025.csv",
    "target_key":"source/nasmatch-portfolio.xlsx",
    "output_key":"output/nasmatch-portfolio-updated.xlsx"
  }' \
/tmp/out.json && cat /tmp/out.json


# make sure the template exists
aws s3 cp ~/Downloads/nasmatch-portfolio.xlsx s3://baseinfra-excelbucket52f10cb5-wwsunmtzl4nc/source/nasmatch-portfolio.xlsx

# upload CSV (should auto-trigger lambda)
aws s3 cp ~/Downloads/Portfolio_Positions_Aug-21-2025.csv s3://baseinfra-excelbucket52f10cb5-wwsunmtzl4nc/source/Portfolio_Positions_Aug-21-2025.csv

aws s3 cp data/positions-2025-08-23.csv s3://baseinfra-excelbucket52f10cb5-wwsunmtzl4nc/source/positions-2025-08-23.csv

# watch logs
aws logs tail /aws/lambda/BaseInfra-ExcelProcessorFnB7B43EA3-fRqa6Yu3FXTU --follow

# verify output
aws s3 cp s3://baseinfra-excelbucket52f10cb5-wwsunmtzl4nc/output/nasmatch-portfolio-updated.xlsx ./nasmatch-portfolio-updated.xlsx