Skip to content

BOS

An agent out of the box. A framework for your own agent-native applications.

BOS gives you a working agent in one command — then grows with you into your own agent-native application, from a single assistant to a multi-agent system.

Quick Start

OPENAI_API_KEY=<api-key> uvx boscli ask "how are you" --model openai/gpt-4o

The CLI ships as the boscli distribution. pip install bos-ai installs the library and provides no boscli command — see Embedding BOS.

Prefer to set the credentials and model once instead of repeating them on every call? Export them into your environment — boscli ask reads BOS_MODEL when --model is omitted:

export OPENAI_API_KEY=<api-key>
export BOS_MODEL=openai/gpt-4o
boscli ask "how are you"

Using a different provider? See LiteLLM's provider docs for the right BOS_MODEL prefix and required environment variables. For example, with a DeepSeek model:

export DEEPSEEK_API_KEY=<api-key>
export BOS_MODEL=deepseek/deepseek-v4-pro
boscli ask "how are you"

Install uv to get uvx, or install the CLI permanently with uv tool install boscli / pipx install boscli:

export OPENAI_API_KEY=<api-key>
export BOS_MODEL=openai/gpt-4o
boscli ask "how are you"

Where to go next

  • Getting Started — set up a project and run the agent runtime.
  • Tutorials — a hands-on path from your first agent to a packaged, shareable extension.
  • Concepts — how the runtime, agents, actors, memory, and skills fit together.
  • Configuration — the complete config.toml reference.
  • Extending BOS — write tools, plugins, channels, and providers.
  • Embedding BOS — put the agent inside your own application: call it directly, or mount the whole gateway runtime.
  • CLI — the boscli commands you'll use day to day.
  • Architecture — the design and principles behind BOS, plus an index of the enhancement proposals (BEPs) that record deeper design decisions.

Building on BOS with an AI agent?

Every project scaffolded by boscli init includes an llm-full.md — a single dense reference that covers every mechanism (configuration, extension points, plugins, channels, skills, the CLI, and the runtime) in one file, intended for ingestion by an AI coding agent. Source: src/bos/llm-full.md.