Quickstart
This guide assumes Raw is installed. If not, follow Install Raw first.
-
Create the starter configuration.
Terminal window raw config initThis writes
~/.config/raw/config.jsonwith a default agent namedraw. It does not overwrite an existing file. -
Point the model at a real model.
Open the file and find the
modelsblock. The starter entry looks like this:{"default_agent": "raw","models": {"local": {"provider": "ollama","method": "openai-chat-completions","model_id": "YOUR_INSTALLED_MODEL","base_url": "http://127.0.0.1:11434/v1"}}}Replace
YOUR_INSTALLED_MODELwith the exact model ID your server reports. For a hosted model, add another entry tomodelsinstead. Keep the API key in an environment variable and reference it withapi_key_env:{"models": {"flash": {"provider": "deepseek","method": "openai-chat-completions","model_id": "deepseek-flash","base_url": "https://api.deepseek.com","api_key_env": "DEEPSEEK_API_KEY","context_window_tokens": 1048576}}}Then point an agent at the new alias by setting its
modelfield toflash. See Models and agents for every field. -
Run a first task.
Terminal window raw "Explain the tests in this repository"Raw runs the task with the default agent, prints the answer to stdout, and writes status lines to stderr. When the run ends, it prints a copyable resume command. To choose an agent explicitly, use
raw --agent raw "task". -
Open the interactive REPL.
Terminal window rawThe REPL keeps one saved session. Type a task at the
❯prompt. Use/compactto summarize older context,/clearto start a new saved session,/statsto show usage, and/exitto leave. Press Ctrl-C to cancel active work. -
Continue the last session.
Terminal window raw --continue "Follow up on that change"--continueresumes the most recent unexpired session for the current workspace. To resume a specific session, useraw --resume ID "task". To list sessions, runraw sessions. -
Open the dashboard.
Terminal window raw dashboardRaw starts a local server at
http://127.0.0.1:8787, opens your browser, and prints an authenticated link. Keep the process running while you use the dashboard. Setup and saved history work even before the model connection is configured.
Next steps
Section titled “Next steps”- Core concepts explains how models, agents, tools and skills fit together.
- Permissions explains what a tool call can do on your machine.
- Dashboard describes the browser app.