Skip to content

Intro

AlphaBase product overview and getting started flow.

AlphaBase is an AI engineering workspace with one local runtime and multiple app surfaces:

  • CLI + TUI (packages/alphabase)
  • Desktop app (packages/desktop)
  • VS Code extension (packages/extensions/vscode-v2)
  • Browser UI (alphabase web, backed by packages/app)
  • API/SDK integrations (packages/sdk/js)

The clients use the same runtime contracts for workspace state, provider connections, sessions, tools, and task output. The available behavior depends on the client version, local configuration, connected provider, model, and permissions.

Start In 10 Minutes

  1. Run AlphaBase from a source checkout (the currently verified path):
Terminal window
git clone https://github.com/XtremeHarness/alphabase.git
cd alphabase
bun install
bun dev /path/to/project
  1. If you are already in the repository, open your project directory with:
Terminal window
cd /path/to/project
  1. Launch the TUI from the project root:
Terminal window
alphabase
  1. Connect at least one model provider.
  2. Pick a model and the active agent/harness.
  3. Run your first prompt:
Create hello-world.txt with one line: Hello, AlphaBase.

Choose Your Surface

  • Use CLI for automation and terminal workflows.
  • Use TUI for interactive terminal-first work.
  • Use IDE for editor-native chat and tool execution.
  • Use Web for browser-based local sessions.

Core Runtime Concepts

  • Providers: model backends and credentials.
  • Tools: file/system/tool execution contracts.
  • Server: local runtime API used by clients.
  • Troubleshooting: health checks and failure recovery.
  • Ecosystem: package map and deployment topology.