# Getting Started This page takes you through five minutes of setup: install, connect an LLM provider, open the interfaces, send your first message, and install your first ready-made agent program. ## Step 1: Install The one-command install script sets up the Python package, web UI, terminal UI, browser tools, and channels: ```bash git clone https://github.com/Fzkuji/OpenProgram.git && cd OpenProgram ./scripts/install.sh # macOS/Linux · Windows: .\scripts\install.ps1 ``` Requires Python ≥ 3.11, Node ≥ 20, and git (the script makes a best effort to install what's missing). The script is idempotent — re-run it any time. Agent programs (GUI / Research / Wiki) are not installed by default; pick them in the interactive menu during install, or add them later (see Step 5). Full flags and the dependency matrix: [Install](../install/install.md). ## Step 2: First run — connect a provider ```bash openprogram ``` The first run enters a setup wizard that walks you through provider configuration — import credentials from a logged-in Claude Code / Codex / Gemini CLI, or paste an API key — then drops you straight into the terminal chat. Re-run the wizard any time with `openprogram setup`. You can also skip the wizard with environment variables: ```bash export ANTHROPIC_API_KEY=sk-ant-... # Claude export OPENAI_API_KEY=sk-... # GPT export GEMINI_API_KEY=... # Gemini (GOOGLE_API_KEY also works) ``` Sanity check: `openprogram providers` lists the detected credentials. ## Step 3: Open the web UI ```bash openprogram web ``` This starts the background worker and opens your browser at **http://localhost:18100** — a single port serving the web UI, the API, and the WebSocket. Change it with `openprogram ports --port
`.
## Step 4: Send your first message
Type directly into the terminal chat or the web input box. To quickly verify from the command line:
```bash
openprogram --print "Introduce yourself in one sentence"
```
It sends one message, prints the reply, and exits. Resume an earlier session with `openprogram --resume