Sobre a vaga
🚀 Full Stack Software Engineer — Agentic Development & MCP Integrations (Claude)
📍 Remote | 🌎 Fluent English (US-based stakeholders) | Level: Senior
**About the role**
We build software from scratch using a multi-agent development pipeline powered by
**Claude**
(Claude Code / Anthropic API), with a
**seven-agent architecture**
covering specification, planning, coding, optimization and review.
You'll be the
**human-in-the-loop**
: orchestrating and steering the agent pipeline, designing and maintaining
**MCP (Model Context Protocol)**
integrations with each client's chosen systems, and acting as the final authority on code quality, stability and production readiness.
A hands-on senior role that combines full stack engineering, QA ownership and AI-agent orchestration.
**What you'll do**
🤖
**Agent pipeline orchestration**
- Operate and supervise the seven-agent software factory (architect, spec-designer, manual-tester, tech-planner, coder, optimizer, reviewer), acting as approver at each human gate.
- Write, refine and version agent definitions, prompts, skills and guardrails; measure output quality and token cost, and iterate.
- Decide when agents proceed autonomously and when to intervene, re-plan or hand-code.
🔌
**MCP architecture & integrations**
- Design, build and secure MCP servers/clients that connect solutions to any client system (CRMs, ERPs, data platforms, ticketing, internal APIs).
- Define tool schemas, authentication (OAuth / API keys / secrets management), permissions scoping, rate limiting and audit logging.
- Keep integrations client-agnostic and reusable: a pattern per system type, not a one-off per customer.
💻
**Full stack delivery**
- Own end-to-end architecture: frontend, backend/APIs, data layer, IaC and CI/CD.
- Review, refactor and complete agent-generated code so it meets production standards.
✅
**Quality assurance & code stability**
- Define acceptance criteria and test strategy per feature before agents start work (spec gate).
- Review all agent output for correctness, security, performance and maintainability; run and extend automated tests (unit, integration, e2e, contract tests for MCP tools).
- Validate agent-generated tests themselves — no self-certified code ships.
- Sign off on releases; maintain regression suites and observability (logging, tracing, error budgets) for both the applications and the agent runs.
**Required qualifications**
- 6+ years of full stack development (e.g. TypeScript/Node + React, or Python + a modern frontend), with production ownership.
- Hands-on experience building with
**Claude**
: Claude Code, Anthropic Messages API, tool use/function calling, prompt and skill engineering.
- Practical experience with
**MCP**
: building/deploying MCP servers, defining tools/resources, and integrating third-party systems.
- Strong
**QA**
background: test automation, code review at scale, security-minded (OWASP, secrets handling, input validation).
- Cloud and DevOps fluency (AWS/GCP/Azure, containers, CI/CD, IaC).
- Ability to read and critique large volumes of generated code quickly and spot subtle defects.
- **Fluent English**
, comfortable working directly with US-based stakeholders.
**Preferred qualifications**
- Multi-agent frameworks / orchestration patterns (planner/executor/critic, human-in-the-loop gates).
- LLM evaluation (evals, golden datasets, regression testing of prompts/agents) and token/cost management.
- Jira/Confluence-driven delivery and spec-first methodologies.
- Prior work in a consulting or nearshore delivery model.