> ## Documentation Index
> Fetch the complete documentation index at: https://docs.orcapods.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Start in your workspace

> Run Orcacode from the workspace where you want its file tools to operate. It defaults to a local OpenAI-compatible endpoint when no API key is set.

## Install Orcacode

On macOS and Linux, one command installs the binary:

```bash theme={"dark"}
$ curl -fsSL https://orcapods.ai/orcacode.sh | sh
```

The installer detects your OS and architecture, downloads the matching release, verifies its SHA-256 checksum, and writes a single executable to `~/.local/bin`. There is no runtime to install alongside it. If that directory is not on your `PATH`, the installer tells you how to add it.

On Windows, run the PowerShell installer:

```bash theme={"dark"}
> irm https://orcapods.ai/orcacode.ps1 | iex
```

Every build is also attached to the [GitHub releases page](https://github.com/okikorg/orca-harness/releases) with a `SHA256SUMS` file. `orcacode update` installs the latest release in place.

| Environment variable | Use it when |
| - | - |
| `ORCACODE_VERSION` | You want to pin a release tag, for example `orcacode-v0.7.0`, instead of taking the latest. |
| `ORCACODE_INSTALL_DIR` | You want the binary somewhere other than `~/.local/bin`. |

To work on the harness itself rather than use it, build from a checkout with `cargo run --release -p orcacode`.

## Your first prompt

Run `orcacode` from the workspace you want its file tools rooted in, or pass `--workspace DIR` to point them elsewhere.

```bash theme={"dark"}
$ orcacode

› Read the README and explain how this project is structured.
› Now identify the smallest test command for the CLI.
```

Use `/help` to open the command list. Type `/` to filter it. Press `Esc` to interrupt a running turn; prompts sent during a run are queued in order.

## Useful launch options

| Option | Use it when |
| - | - |
| `--workspace DIR` | Tools should be rooted in another directory. |
| `--model NAME` | You know the model identifier to use. |
| `--base-url URL` | Your compatible endpoint is not the default. |
| `--openrouter` | You want OpenRouter as the active endpoint. |
| `--plan` | You want a read-first planning session. |
| `--normal` | You want to state the default human-approval mode explicitly. |
| `--auto` | You want guarded edits to proceed automatically and remaining risks reviewed. |
| `--orchestrate` | You want the agent to delegate substantial work while permitting basic direct edits. |
| `--continue` | You want the latest session for this workspace. |

<Note>
  **Check before you act:** Normal mode is the default and asks before gated actions in the interactive terminal. `--auto` replaces those prompts with automatic admission and review; `--yolo` removes review entirely. Use plan mode when no state-changing action should run.
</Note>


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.