Senior Engineer Agent

A LangChain/LangGraph agent that codes like a senior software engineer. Install it globally once, use it in any project — no agent code in your repo.

Quick Start

# 1. Install globally
npm install -g senior-engineer-agent

# 2. Configure your API key (one-time)
senior-engineer setup

# 3. Add the agent to any Kiro workspace
cd your-project
senior-engineer init

# 4. Use it — from CLI or Kiro
senior-engineer "Write a Python LRU cache with O(1) get and put"

That's it. Open the project in Kiro and invoke /senior-engineer from chat.

How It Works

The agent runs from your global npm install. Your project only gets a thin .kiro/agents/senior-engineer.md file (created by senior-engineer init) that tells Kiro how to invoke the globally installed CLI. No agent source code is downloaded into your project.

┌─────────────────────────────────────────────────┐
│  Your Project                                   │
│  .kiro/agents/senior-engineer.md  (tiny file)   │
│       │                                         │
│       ▼  invokes                                │
│  senior-engineer CLI  (global npm package)      │
│       │                                         │
│       ▼  calls                                  │
│  Groq / OpenAI API                              │
└─────────────────────────────────────────────────┘

Architecture

The agent uses a LangGraph state graph with six nodes:

analyze_task ──► plan_approach ──► generate_code ──► generate_tests
                                                          │
                                                          ▼
                                                    review_code
                                                     │       │
                                          (approved) │       │ (needs_refactor)
                                                     ▼       ▼
                                                    END   refactor_code ──► generate_tests
                                                             │                    ...
                                                             ▼ (max iterations)
                                                            END

| Node | Purpose | |------|---------| | analyzeTask | Extracts language, complexity, requirements, and constraints | | planApproach | Creates architecture plan, implementation steps, design patterns | | generateCode | Writes production-quality code following the plan | | generateTests | Generates comprehensive tests | | reviewCode | Reviews for bugs, security, performance, and style | | refactorCode | Addresses review issues and loops back for re-testing |

CLI Reference

# Run the agent on a task
senior-engineer "Build a REST API rate limiter in Go"

# With options
senior-engineer --max-iterations 5 --verbose "Implement a B-tree in Rust"

# Pipe from stdin
echo "Write a binary search in TypeScript" | senior-engineer

# Setup & init
senior-engineer setup              # Configure API key (one-time)
senior-engineer init               # Add agent to current Kiro workspace
senior-engineer --help             # Show help

Node.js API

import { run } from "senior-engineer-agent";

const result = await run(
  "Write a thread-safe singleton pattern in Java",
  { maxIterations: 3 },
);

console.log(result.code);
console.log(result.tests);
console.log(result.reviewVerdict); // "approved" or "needs_refactor"

Configuration

Config is loaded in this order (first wins):

1. Environment variables
2. Local .env in current directory
3. Global ~/.senior-engineer/.env (created by senior-engineer setup)

| Variable | Default | Description | |----------|---------|-------------| | LLM_PROVIDER | groq | groq (free) or openai | | GROQ_API_KEY | — | Free at console.groq.com/keys | | OPENAI_API_KEY | — | Only if using openai provider | | LLM_MODEL | provider default | Model name override | | LLM_TEMPERATURE | 0.2 | Sampling temperature |

Available Groq Models (free)

| Model | Context | Best for | |-------|---------|----------| | llama-3.3-70b-versatile (default) | 128k | Complex coding tasks | | llama-3.1-8b-instant | 128k | Fast, simple tasks | | gemma2-9b-it | 8k | Balanced performance | | mixtral-8x7b-32768 | 32k | Good reasoning |

For Other Developers

Any developer on your team can start using the agent in 3 commands:

npm install -g senior-engineer-agent
senior-engineer setup
senior-engineer init

No agent code ends up in the project. The .kiro/agents/senior-engineer.md file is safe to commit — it just tells Kiro to call the global CLI.

Testing

# Clone the repo for development
git clone <repo-url> && cd senior-engineer-agent
npm install

# Run tests
npm test

# Type checking
npm run typecheck

License

MIT