harness

Install

curl -fsSL /install.sh | sh

Linux (x86_64) and macOS 15.4 or later (Apple Silicon or Intel). Installs one binary to ~/.local/bin, no sudo. Read the script before you run it.

On Linux the sandbox needs bubblewrap: sudo apt install bubblewrap or your distribution's package.

To update, run the install command again. To uninstall, delete ~/.local/bin/harness.

No key is needed to start: run harness run, then type /login.

Two ways to start

Free models

Run harness run and type /login (or harness login in the shell). Open the link it prints, check the code matches, and sign in with Google.

You get mega:flash (fast and cheap) and mega:full (smartest, slower), free up to a daily allowance that resets at 00:00 UTC. mega:full uses it about 2.7× as fast. harness logout signs out. Privacy.

Your own keys

Add these to your shell's startup file (~/.zshrc or ~/.bashrc), then open a new terminal.

1. TypeSafe

export TYPESAFE_API_KEY=your-key

Needed for auto mode with your own keys. Signed in, the judge runs through the free service. TypeSafe judges each command before auto mode runs it. Get a key at typesafe.ai.

2. Your model provider

export ANTHROPIC_API_KEY=your-key   # or any variable below
ProviderVariable
AnthropicANTHROPIC_API_KEY
OpenAIOPENAI_API_KEY
GeminiGEMINI_API_KEY
OpenRouterOPENROUTER_API_KEY
TogetherTOGETHER_API_KEY
DeepSeekDEEPSEEK_API_KEY
Z.AIZAI_API_KEY
xAIXAI_API_KEY
Ollama, LM Studio (local)no key

Start

harness run                                   # signed in: mega:full; else your first provider with a key
harness run --model gemini:gemini-3.8-flash   # any provider: --model provider:model

To start on another model every time, put it in ~/.config/harness/config.toml:

model = "gemini:gemini-3.8-flash"

Rust SDK

Run harness sessions from your own program and give the model tools you wrote.

cargo add megacode-agent-sdk

Guide · crates.io · API reference

Docs

The docs cover modes, the sandbox, /undo, shortcuts and every setting.