Aztec

Privacy-first, local-execution coding agent

Quick Start

Option 1: Using Ollama (Recommended for Local Execution)

# 1. Install Ollama from https://ollama.com
# 2. Pull the model
ollama pull qwen2.5-coder:7b

# 3. Create custom Aztec model
ollama create aztec -f Modelfile

# 4. Run Aztec
./aztec.exe -m aztec

Option 2: Direct Ollama Model

ollama pull qwen2.5-coder:7b
./aztec.exe -m qwen2.5-coder:7b

Option 3: Cloud Providers

# OpenAI
./aztec.exe --provider openai -m gpt-4o

# Universal (any OpenAI-compatible API)
./aztec.exe --provider universal -m google/gemini-2.0-flash

---

System Requirements

Component	Minimum	Recommended
CPU	Core i5 8th gen	Core i7 / Ryzen 7
RAM	8GB	16GB+
Storage	10GB free	SSD
GPU	Not required	NVIDIA for speed

For Ollama + Qwen2.5-Coder-7B:

• Model size: 4.3GB
• RAM usage: ~6-8GB while running
• Speed: ~5-10 tokens/second on Core i5

---

Core Features

Tool Ecosystem

Aztec executes real tools to get work done:

Tool	Description
TERMINAL	Run shell commands (mkdir, npm install, go build, etc.)
WRITE_FILE	Create new files with complete code
READ_FILE	Read existing files
EDIT	Modify existing files (search/replace)
LIST_FILES	Explore directory structure
GLOB	Find files by pattern
GREP	Search file contents
TODO_WRITE	Track progress on complex tasks
DONE	Signal task completion
GIT_*	Git operations (diff, status, commit, branch)

CLI Flags

aztec [OPTIONS]

OPTIONS:
  -w, --workspace <DIR>    Set workspace directory
  -m, --model <NAME>       Set model name
  -p, --provider <TYPE>    Set provider (ollama, openai, vllm, lmstudio, universal)
  --approval <MODE>        Set approval mode (confirm, auto, yolo)
  -v, --verbose            Enable verbose output
  --no-color               Disable colored output
  -h, --help               Print help
  --version                Print version

SESSION COMMANDS:
  --list-sessions          List saved sessions
  --resume <id>            Resume a session
  --fork <id>              Fork a session
  --export <id>            Export session to markdown