You will get AI agents and workflow automation for your back-office, MCP-connected
Project details
You get autonomous AI agents that do real, repetitive back-office work, invoice reconciliation, ticket triage, CRM hygiene, recurring reports, and run on a schedule or an event with every step inspectable. Not a chatbot wrapper: you watch a run unfold step by step, see each tool-call and its result, read the logs live, and step in when the agent pauses for review.
The stack is Claude and OpenAI for reasoning, MCP to connect agents to your SaaS tools and internal APIs, TypeScript or Python for orchestration, and a queue plus scheduler for runs that survive restarts and retries. Several tools in sequence, or several agents handing work to each other, stay observable: ordered steps, typed arguments, results, and a streamed log per run.
I built a live demo so you can judge it first: a control room with a live run, a workflow library, and an MCP integrations surface, a fictional company rather than client work. The same approach runs in my own products: a self-hosted AI tool that ships its own MCP server, and an engine that drives a coding agent in a loop.
I scope tightly, give a fixed price up front, and hand over code you own outright.
The stack is Claude and OpenAI for reasoning, MCP to connect agents to your SaaS tools and internal APIs, TypeScript or Python for orchestration, and a queue plus scheduler for runs that survive restarts and retries. Several tools in sequence, or several agents handing work to each other, stay observable: ordered steps, typed arguments, results, and a streamed log per run.
I built a live demo so you can judge it first: a control room with a live run, a workflow library, and an MCP integrations surface, a fictional company rather than client work. The same approach runs in my own products: a self-hosted AI tool that ships its own MCP server, and an engine that drives a coding agent in a loop.
I scope tightly, give a fixed price up front, and hand over code you own outright.
AI Algorithms
Large Language Model, Transformer ModelAI Applications
AIOps, Natural Language Generation, Natural Language UnderstandingAI Models
ChatGPT, GPT-4What's included
| Service Tiers |
Starter
$3,000
|
Standard
$7,000
|
Advanced
$15,000
|
|---|---|---|---|
| Delivery Time | 7 days | 14 days | 24 days |
Number of Revisions | 2 | 2 | 3 |
AI Model Integration | |||
Batch Normalization | - | - | - |
Database Integration | - | ||
Detailed Code Comments | |||
Image Upscaling | - | - | - |
MLOps | - | - | - |
Model Deployment | |||
Model Documentation | |||
Model Monitoring | - | - | |
Model Testing & Optimization | - | - | |
Model Tuning | - | - | - |
Natural Language Processing | - | - | - |
NLP Tokenization | - | - | - |
Pre-Training | - | - | - |
Prompt Engineering | |||
Setup File | - | - | - |
Source Code |
Optional add-ons
You can add these on the next page.
Additional MCP Integration
(+ 3 Days)
+$1,500
RAG Over Your Documents
(+ 5 Days)
+$2,500
Evaluation & Guardrails
(+ 5 Days)
+$2,000Frequently asked questions
About Miguel
AI-Native Product Engineer - Chatbots, Agents & SaaS MVPs
Barcelona, Spain - 4:59 am local time
Recent work shows what that looks like in production. I built an AI-powered education platform on Next.js and Expo using GPT-5 and Claude, including an SDK that was adopted across multiple institutions. I built a fintech digital wallet in React Native and Expo, backed by a NestJS and TypeScript microservices stack on AWS. And I ship my own AI products: a self-hosted AI pipeline workbench with its own MCP server, an agent-native engine that runs agents in a loop, a conversational booking app, and AI editorial and design tools.
What that means for you: I understand the full path from a model call to a deployed, reliable product. Auth, billing, retrieval, tool-calling, evaluation, and the unglamorous parts (logging, retries, guardrails) that decide whether an AI feature survives contact with real users.
How I like to work: tight scope, clear milestones, and a working demo early so you can steer before the budget is spent. I write clean, typed, tested code and I document what I hand over.
If you want a defined outcome at a fixed price, I package the most common builds (chatbots, RAG assistants, agents, SaaS MVPs, AI-commerce storefronts) as fixed-price projects below. Start there, or message me to scope a custom build.
Steps for completing your project
After purchasing the project, send requirements so Miguel can start the project.
Delivery time starts when Miguel receives requirements from you.
Miguel works on your project following the steps below.
Revisions may occur after the delivery date.
Scope and integrations
We lock the task and rules, then connect the agent to your tools and APIs via MCP, with credentials in place.
Build the run loop
I build the agent: reasoning, tool-calls, retries, logging, and a run console you can watch step by step.