Skip to main content
TripleBits Apprentice - Software for Agents

Menu

Sign In Register

Getting Started with Apprentice

A short first-run guide for installing prerequisites, choosing a provider, creating a focused first agent, and reviewing its first run.

35 views

Getting Started with Apprentice

Apprentice is a desktop app for creating, running, and monitoring AI agents from your own computer. It is not a hosted SaaS where your agents run on Triple Bits servers.

Apprentice is local-first, but it is not local-model-only. The app, agent configuration, runtime decisions, logs, memory, tasks, schedules, and audit history are managed locally. Model calls go to whichever provider or local runtime you choose.

What You Need

Before creating your first agent, make sure you have:

  • Apprentice installed on Windows, macOS, or Linux.
  • Docker installed and running.
  • At least one AI provider or local model runtime configured.
  • Optional: a folder or project you want an agent to work with.

Docker is required because Apprentice uses a Docker-backed runtime to run agents in a controlled Linux environment instead of running agent commands directly on your host system.

Start with these guides:

  1. Install Docker for Apprentice.
  2. Connect Your First AI Provider.
  3. Create Your First Agent.
  4. How Local-First Works in Apprentice.
  5. Run And Review Your First Agent.

You only need one working provider and one simple agent to begin. Add more capabilities later.

Choose a Provider

Apprentice can work with different provider types:

  • API providers such as OpenAI, Anthropic / Claude API, Google Gemini API, DeepSeek, Mistral, Kimi, GLM, and Qwen.
  • CLI providers such as Claude Code, Codex, and Gemini CLI.
  • Subscription or OAuth paths such as ChatGPT Subscription.
  • Local model runtimes such as LM Studio, Ollama, and Docker Model Runner.

If you use a local runtime, model calls can stay local. If you use an API, CLI, or subscription provider, that provider receives the prompt and context needed for the model call.

Create a Small First Agent

For your first agent, keep the scope narrow.

A good first agent might be:

Help me organize tasks, summarize notes, and answer questions about one project folder.

Start with:

  • One provider.
  • One model.
  • One clear role.
  • Limited permissions.
  • A small budget or run limit.
  • Folder access only if the agent actually needs it.

Avoid starting with broad filesystem access, many integrations, browser automation, and MCP tools all at once. Apprentice is designed so you can expand an agent's capabilities gradually.

Run a First Message

Try a simple request:

Summarize what you can help me with based on your current configuration.

If you gave the agent access to a folder, try:

Inspect this project folder and summarize the main areas.

Review the Run

After the run, check what happened:

  • Which provider and model were used.
  • Whether any permissions were requested.
  • Which files, tools, or commands were involved.
  • Whether the result matched the agent's role.
  • Whether the agent needs more or fewer permissions.

That is the core Apprentice workflow: configure an agent, run it, inspect what happened, then adjust its boundaries.