You will get a Custom MCP Server with FastMCP — Claude, Cursor & AI-Ready Tools

Najib M.Status: Offline
Najib M. Najib M.

Let a pro handle the details

Buy Other AI & Machine Learning services from Najib, priced and ready to go.
Najib M.Status: Offline
Najib M. Najib M.

Let a pro handle the details

Buy Other AI & Machine Learning services from Najib, priced and ready to go.

Project details

Are you building AI workflows with Claude, Cursor, or any MCP-compatible client and need custom tools that actually do things — search the web, query your database, call your APIs, send messages, manage files?

I build production-ready MCP servers using FastMCP and Python. You describe the tools you need. I architect, build, test, and deploy them. Your AI goes from answering questions to taking real actions.

I've shipped a 19-tool MCP server in production — covering web search, database reads/writes, Redis caching, file operations, external API calls, and LLM-powered summaries. This isn't a template — every server is built to your exact use case.

What you get: Clean Python codebase, full schema documentation, deployment to Railway or Render, and a Claude Desktop config file ready to drop in and use.

Message me before purchasing if you have a custom scope — I'm happy to discuss.
AI Development Type
Knowledge Representation, Software Maintenance
AI Development Language
Python
What's included
Service Tiers Starter
$800
Standard
$2,500
Advanced
$4,500
Delivery Time 7 days 12 days 25 days
Number of Revisions
123
AI Model Integration
Detailed Code Comments
Knowledge Graph
-
-
-
Model Documentation
Ontology
-
-
-
Source Code
Taxonomy
-
-
-
Optional add-ons You can add these on the next page.
Additional Revision
+$200

Frequently asked questions

Najib M.Status: Offline
Najib M.Status: Offline
AI Agent & Custom GPT Developer - Python - LLM - On-Chain Analytics
Singapore, Singapore - 4:14 am local time
Most AI freelancers build demos. I build systems that real people use, that run without supervision, and that solve actual problems.
My background spans high-frequency trading infrastructure, on-chain analytics, and production Custom GPT development. That combination is unusual — and it means I can handle both the heavyweight technical builds and the practical business automation work that most clients actually need.
━━━━━━━━━━━━━━━━━━━━━━

WHAT I BUILD

━━━━━━━━━━━━━━━━━━━━━━
→ Custom GPT systems with real backends — Supabase, Google Sheets, OAuth gates, API integrations, user capture, and structured workflows. Not chatbots. Operational tools.

→ AI agents with tool use, memory, and multi-step reasoning using LangChain, LlamaIndex, and the Claude/OpenAI/Mistral APIs

→ RAG pipelines over proprietary data with vector databases (Pinecone, pgvector, FAISS)

→ MCP (Model Context Protocol) servers — giving AI systems direct programmatic access to your APIs, databases, and external data sources

→ Real-time data pipelines feeding into LLM analysis and automated decision layers

→ Market intelligence dashboards with AI-generated signal detection and commentary

→ Autonomous decision engines with multi-gate filtering and systematic execution logic

→ Browser extensions with Python backend services for live data capture and processing
━━━━━━━━━━━━━━━━━━━━━━

WHAT I HAVE ACTUALLY SHIPPED

━━━━━━━━━━━━━━━━━━━━━━
On the Custom GPT side:
A tournament intelligence platform for the 2026 FIFA World Cup — fan discussion engine covering team analysis, match context, odds sentiment, and prediction tracking, with email gating, Supabase backend, user prediction storage, and an OAuth access layer. Built to feel alive during a live tournament, not like a static FAQ.
A booking and lead generation assistant for a Singapore-based children's events business — face painting, balloon sculpting, corporate family days, birthday parties. Live and in use. Handles lead qualification, corporate targeting, outreach support, and contact logging via Google Sheets with live handover to the business owner.
An AI education product built around the principle of learning by doing — structured 7-day and 30-day programs, interactive practice sessions, progress tracking, artifact creation, and student confidence capture. Backend validation via Google Sheets and Apps Script.
A PSLE exam preparation tool for Singapore P6 students covering Science, Math, and English — instant marking, targeted explanation loops, and a teacher validation workflow with response capture.
On the infrastructure side:
A live wallet intelligence platform monitoring blockchain activity across thousands of addresses in real-time — proprietary scoring algorithm pulling from multiple API sources in parallel, automated trader-style commentary via fine-tuned LLM, coordinated pattern detection, and a structured database of 1,000+ documented entities. Zero manual intervention in the core loop.
A prediction market copy-trading system with on-chain signal generation, a three-gate decision engine (EV filter, fractional Kelly sizing, risk gate), live terminal dashboard, and paper trading mode. Built on systematic trading firm principles.
A custom MCP server with 19 tools across five functional domains, built on FastMCP with full Pydantic v2 validation, running as a persistent AI workflow layer inside Claude Desktop.
━━━━━━━━━━━━━━━━━━━━━━

TECH STACK

━━━━━━━━━━━━━━━━━━━━━━
Python · Claude API · OpenAI API · Mistral · LangChain · LlamaIndex · FastMCP · FastAPI · Supabase · PostgreSQL · Redis · Google Sheets API · Apps Script · WebSocket streaming · Docker · n8n · REST/GraphQL · Browser extension development · Real-time data pipelines
━━━━━━━━━━━━━━━━━━━━━━

HOW I WORK

━━━━━━━━━━━━━━━━━━━━━━
Scope and architecture before any code. I will tell you when a simpler solution exists. I will push back if the approach is wrong. I do not build things that work in demos and break in the real world.
I work best with founders and operators who know what outcome they want and need a developer who can design and build the system to get there.
Fixed-price scoped projects preferred. Clean handover with documentation.
Background in Computer Science and Economics. Singapore-based (UTC+8). Available for async collaboration globally
Message me with what you are trying to build and what it needs to do.

Steps for completing your project

After purchasing the project, send requirements so Najib can start the project.

Delivery time starts when Najib receives requirements from you.

Najib works on your project following the steps below.

Revisions may occur after the delivery date.

Scoping

We align on exactly which tools you need, what data sources they touch, and your MCP client setup. I'll flag anything that affects architecture before a line of code is written.

Build

Each tool is built with full input validation, error handling, and clean response schemas. Every tool is testable in isolation before integration.

Review the work, release payment, and leave feedback to Najib.