Intelligent Programs Powered by Local AI

We build autonomous, AI-powered software — "gods in the machine" — that operate entirely on local, private, offline AI models. No cloud dependencies. No API keys to external services. Just raw local inference powering intelligent programs.

Inspired by the Gods from the Machine raid in Star Wars: The Old Republic — a pantheon of ancient, sentient super-weapons worshiped as deities on the mechanical Dyson sphere planet Iokath.

Explore Projects View on GitHub
Projects

The Pantheon

Each god is a specialized autonomous entity — named after gods and legends from world mythology. Every project runs on consumer hardware using small, capable models served via llama.cpp.

gilgamesh Go

TDD-driven local AI coding agent with CLI, MCP, and HTTP API interfaces. Tool-calling, streaming SSE, skills, hooks, session logging.

ACTIVE — v0.6.0
zeus Zig

Zig-as-a-build-system for llama.cpp-powered local GGUF inference. Build, quantize, and serve models with a single command.

INCOMPLETE
raijin Rust

ONNX CPU inference engine. Local model execution in pure Rust.

INCOMPLETE
garuda Zig

Directory tree viewer replicating PowerShell's tree command. Fast, colorful, cross-platform.

INCOMPLETE
Philosophy

Principles

Local-first

All intelligence comes from free, local, private AI models. No cloud, no subscriptions, no data leaving your machine.

Blazing fast

Built in Go, Rust, Zig — high-performance languages for CPU inference. First response in seconds, not minutes.

Lean

Minimal token overhead. Gilgamesh uses ~1,600 tokens total, enabling responsive agent loops on CPU.

Open source

MIT licensed. Free to use, learn from, and contribute to. No proprietary lock-in.

CLI + MCP + API

Every god exposes the same capabilities through CLI, MCP, and HTTP — for humans, agents, and programs alike.

Mythological

Projects named after gods and legends. Each one a specialized, autonomous entity in the pantheon.

Research

Model Trialing

We run controlled model trials on consumer hardware — no GPU, just CPU inference. Finding the best model + quantization + parameters for a responsive, reliable, tool-calling coding agent.

See Research & Findings

Get Started

Quick Start

terminal
# Clone and build
$ git clone https://github.com/godsfromthemachine/gilgamesh
$ cd gilgamesh && go build -o gilgamesh .

# Interactive mode
$ ./gilgamesh

# One-shot mode
$ ./gilgamesh run "list all Go files"

# MCP server (for Claude Desktop, other agents)
$ ./gilgamesh mcp

# HTTP API server
$ ./gilgamesh serve -p 7777