agent runner / local first

More agent.
Less machine.

memoria is a focused CLI agent runner with a tiny binary and a deliberately small memory footprint. Bring your Codex subscription, OpenRouter account, or OpenCode Zen/Go setup.

$ curl -fsSL https://memoria.dev/install | sh

placeholder install URL — replace with your release endpoint

$ memoria
❯ explore the project

[activate_skill] explore-codebase
[list_directory] "./"
[bash] "git status --short && find . -maxdepth 2 -type f | sort | sed -n '1,160p'"
~ Thought Exploring the codebase
[read_file] "README.md"
[list_directory] "src"

Project overview
Memoria is a terminal-based AI coding assistant written in Odin. The runnable package is src/; the README’s basic launch command is:
┃gpt-6-luna | 6.8k/272k tokens (2.5%)

01 / footprint

Built for the shell you already live in.

No heavyweight runtime between you and the model. memoria stays out of the way, starts fast, and keeps system resources available for your editor, compiler, and tests.

[01]

Low footprint

A compact process with fewer moving parts and less background overhead.

[02]

Low RAM

Designed to leave memory for the work, even across long-running sessions.

[03]

Tiny binary

One small executable. Easy to download, move, pin, and remove.

02 / in use

A quiet interface for loud problems.

Replace the image below with a screenshot of memoria. The frame is already responsive and will preserve a 16:9 crop.

Placeholder for a screenshot of the memoria CLI

03 / compatible

Your model access. Your choice.

Use the provider you already pay for. memoria keeps the runner consistent while the model backend stays flexible.

Codex subscription

ready

Connect your existing Codex subscription and start a session from the terminal.

OpenRouter

ready

Route across supported models with your existing OpenRouter credentials.

OpenCode Zen / Go

ready

Use memoria with OpenCode's Zen and Go model access paths.

One binary. One prompt. Get to work.

copy install command