Skip to content

Baryo.CLI

A terminal AI agent in Go, running against models on your own machine through Ollama or Docker Model Runner, or against 18 cloud providers if you would rather. Interactive when you want a conversation, scriptable when you want a pipeline.

curl -fsSL https://raw.githubusercontent.com/BaryoDev/Baryo.CLI/main/install.sh | shcopy

The problem

Every AI coding tool wants your code on someone else's server, and twenty dollars a month for the privilege. For a lot of work that is a fine trade. For a client's proprietary codebase under NDA, it is not a trade you are allowed to make.

Baryo runs the model on your machine. Nothing leaves it. No API key, no account, and it still works on a plane.

What it does

Two modes, because a CLI that is only interactive cannot be automated, and one that is only scriptable is miserable to actually use.

bash
# Interactive: pick a model, then talk to it
baryo

# One-shot
baryo --model gemma3 -p "explain docker networking"

# Pipe your code in as context
cat main.go | baryo -p "review this code"

# Headless, tools enabled, machine-readable out. This is the CI-friendly mode.
baryo -p "list all Go files" --yolo --output json
  • Local first: Ollama or Docker Model Runner. Cloud providers are optional.
  • Real tool calling, including in headless mode, so it can do work rather than just describe it.
  • Clean stdout and documented exit codes (0 success, 1 runtime, 2 config), which is what makes it usable in a pipeline.
  • MCP support, skills, plan mode, sub-agents, checkpoints and rewind, worktree isolation, sandboxed execution, RAG retrieval and web search.

Installing

The install script above is the quickest path. Homebrew and Scoop also work:

bash
brew tap BaryoDev/Baryo.CLI https://github.com/BaryoDev/Baryo.CLI
brew install baryo

Status

Current release is v0.12.1, a stabilization release: crash fixes, atomic file writes, tool-loop correctness, dependency security updates. It is pre-1.0 and I am still moving it, so read the changelog before you depend on it.

The full command reference and roadmap on GitHub →

Brewed in the baryo ☕ · Released under the MIT License.