Quickstart

1. Install

macOS / Linux (one line):

curl -fsSL https://contenox.com/install.sh | sh

To also preinstall the local inference daemon (a ~600 MB download, otherwise installed on demand by contenox setup or contenox modeld install):

curl -fsSL https://contenox.com/install.sh | CONTENOX_WITH_MODELD=1 sh

Or download the binary directly from GitHub Releases.

The whole local-first path — install, setup, daemon, model pull, first prompt — in one take:

2. Initialize a workspace

Run this once in the project directory you want Contenox to work in:

contenox init

This creates the workspace marker, writes the default chain and HITL policy presets, and ensures the built-in local backend exists.

3. Pull a local model

For the local-first path, install and start the modeld daemon, then pull a curated GGUF model:

contenox modeld install     # downloads + verifies the daemon; prints the serve command
contenox model pull granite-3.2-2b
contenox doctor

Keep the printed modeld serve command running in another terminal — it is the process that loads models and serves inference.

See the modeld guide (and Architecture) for how the daemon works, capacity planning, remote nodes, and the lease/slot model.

Local models are served by the bundled modeld daemon. With contenox serve running, the Beam UI’s modeld console shows the daemon and lets you load or unload the resident model:

On a fresh install, the first pulled model becomes default-model, and contenox init sets default-provider to local when no provider was already configured.

Run your first prompt:

contenox "hello, what can you do?"

For a persistent chat session:

contenox chat -e

4. Optional editor use

Contenox can also run inside editor or desktop clients that speak ACP. The same chains, model config, tools, and HITL policy are used either way:

• Use from Zed
• Use from VS Code or VSCodium
• Use from JetBrains
• Use from AionUi

Cloud providers

Contenox needs at least one model to work. Pick the option that fits:

If you’re not sure, start with built-in local models — no account or API key needed.

All providers can also be configured from the Beam UI (contenox serve), which stores keys and creates the matching backend for you:

Next steps

• Your first chain — author your own agent in five edits
• Core concepts — how chains, tasks, and tools fit together
• MCP integration — connect external tools