# Get started with aixplain Desktop

**[aixplain Desktop](https://www.aixplain.com/koder#download)** is the aixplain app for your computer. It brings Omni, Koder and Design into one app, along with Scan for security reviews and Agents for building and deploying your own agents. Your sessions and settings stay on your computer, and you choose which model does the work. This article explains what each part does, how to install the app, connect a model and choose permissions, how to update or remove it and what it can't do yet.

> aixplain Desktop is free to download for macOS (Apple silicon and Intel), Windows x64 and Linux x64. You pay only for the models you use. The app is in early releases (version 0.x), so features can change from one release to the next.

> **Note:** Your sessions and files stay on your computer. When you deploy an agent or create an asset you want to share with your workspace, you are working in your workspace on cloud, where your team can use it.

## What is aixplain Desktop?

aixplain Desktop is where you work with aixplain agents on your own machine. You describe what you want in plain language, the agent does the steps, and you follow along, answer its questions and approve anything risky. It installs as **aixplain** and shows up under that name in Applications, Spotlight, Launchpad, the Dock and the Start menu.

### What's inside

- **Omni:** a personal assistant for the work around the code: drafting, planning, documents, numbers and the apps you connect.
- **Koder:** an AI coding agent. It reads your code, makes edits you review line by line, runs your tests and asks before anything risky.
- **Scan:** security review for a repository. It lists likely security bugs worst first, traces how each one is reached and can propose a fix.
- **Design:** landing pages, dashboards, app screens, slide decks, animated intros and vertical video, built as files in your project.
- **Agents:** build, test and deploy your own agents, with Omni or in code. Deploy each one on cloud, where it runs 24/7, or on your own servers.
- **Your phone as a remote:** pair a phone or tablet with your computer to follow chats, answer questions and approve steps when you're away from your desk. See **[Pair your phone or tablet](#pair-your-phone-or-tablet)**.

The same agent also runs in your terminal through the `koder` command. See **[Use aixplain from the terminal](#use-aixplain-from-the-terminal)** below.

## Install aixplain Desktop

### Requirements

- **A supported computer:** macOS on Apple silicon or Intel, Windows x64 or Linux x64 (DEB, RPM or AppImage). The CLI and terminal app also run on Linux arm64.
- **A model to work with:** a free aixplain account (no card needed), a Claude, ChatGPT or Cursor plan you already pay for, your own provider key or a local model.

### Download and install

1. Go to **[aixplain.com/koder](https://www.aixplain.com/koder#download)** and pick the installer for your computer, or download it from the table below.
2. Open the installer. On macOS, open the DMG and move **aixplain** to your Applications folder. On Windows, run the installer. On Linux, install the DEB or RPM package, or make the AppImage executable and run it.
3. Open **aixplain**.

| Your computer | Installer |
| --- | --- |
| macOS, Apple silicon | [Download DMG](https://github.com/aixplain/koder/releases/latest/download/aixplain-code-desktop-mac-arm64.dmg) |
| macOS, Intel | [Download DMG](https://github.com/aixplain/koder/releases/latest/download/aixplain-code-desktop-mac-x64.dmg) |
| Windows x64 | [Download EXE](https://github.com/aixplain/koder/releases/latest/download/aixplain-code-desktop-win-x64.exe) |
| Linux x64 | [DEB](https://github.com/aixplain/koder/releases/latest/download/aixplain-code-desktop-linux-amd64.deb), [RPM](https://github.com/aixplain/koder/releases/latest/download/aixplain-code-desktop-linux-x86_64.rpm) or [AppImage](https://github.com/aixplain/koder/releases/latest/download/aixplain-code-desktop-linux-x86_64.AppImage) |

> **Note:** The macOS app is signed with aixplain's Developer ID and notarized by Apple, and the Windows installer is signed by aixplain. Every release includes a `SHA256SUMS` file so you can verify your download. Earlier versions are on **[GitHub](https://github.com/aixplain/koder/releases)**.

## Sign in

The first time you open the app, it asks you to sign in to aixplain. Your browser opens to finish signing in, and the app also shows the sign-in link with a **Copy** button in case the browser doesn't open. Signing up is free, with no card needed, and new accounts start with free credits. For your account and team, see **[Account, plans and billing](https://www.aixplain.com/support/articles/account-plans-and-billing)**.

## Choose a model

Pick the model that does the work. You can change it at any time from the model picker in the chat box.

| Option | What you need | How you pay |
| --- | --- | --- |
| **aixplain models** | Your aixplain account | Your team's aixplain credits |
| **A plan you already have** | Claude Code, Codex CLI or Cursor's agent, installed and signed in | Counts against that plan's limits |
| **Your own provider** | An API key or a chat-completions-compatible endpoint | Billed by your provider |
| **A local model** | Ollama or the built-in engine | Nothing beyond your hardware |

### Use aixplain models

Sign in to aixplain to use its catalog of hosted models, including open-weight models. New accounts start on Athena, aixplain's meta-model. For every request, Athena chooses how to answer from several models, so you get a fitting model without picking one yourself. If your workspace's catalog doesn't include Athena, the app starts you on GLM 5.3.

The model picker opens on a short list: top coding models, models on a free offer and your recent picks. Each row shows the price per million tokens. Search covers every model from the first letter you type, and hovering a model shows its data retention, limits, price and host.

To see only the models the catalog reports as zero data retention, pick **ZDR only** in the model picker, or run `koder models --zdr` in the terminal.

### Use a plan you already have

Install the vendor's CLI (Claude Code, Codex CLI or Cursor's `cursor-agent`) and sign in to it with your plan's account. The app finds installed CLIs automatically, including ones behind version managers. Check where each plan stands in **Settings → Models & providers → Your subscriptions**. For the details, see **[Get started with Koder](https://www.aixplain.com/support/articles/get-started-with-koder)**.

### Connect your own provider

1. Open **Settings → Models & providers**.
2. Choose the provider and add your API key, or add any chat-completions-compatible endpoint by its base URL.
3. Pick one of its models in the model picker.

> **Note:** If your endpoint uses a self-signed certificate, the app has a toggle for it when you add a custom model.

### Run a model on your computer

A local model runs on your own machine, so inference stays there and doesn't count toward your usage.

1. Start setting up a local model in the app.
2. Pick the suggested model, or another one. The app looks at your computer's memory, graphics and free disk and suggests a model that fits.
3. Wait for the download and setup to finish. The app checks that the model actually answers before it offers it.
4. Pick the local model in the model picker.

If Ollama is already running, the app finds it with no setup and reads which models can chat, call tools and accept images. It can also use a llama-server or LM Studio that is already running on your computer.

> **Note:** Local models need a context of at least 32,768 tokens. Models trained on less are left out when the app routes requests on its own. Sign-in and connected apps still need a network connection.

### Switch models

Open the model picker in the chat box and pick another model. The chat box shows the full name of the model in use and the thinking level the request really uses.

- **Images and PDFs:** if the model you picked can't read an attachment, the app tells you before it sends the message.
- **Another agent:** you can hand a chat to another agent partway through, and the app sizes what it carries over to fit that agent's model.

In the terminal, run `koder models` to list your models, or add `--model <provider/model>` to `koder run` to pick one for a single run.

## Choose how much the agent does on its own

Pick an access mode from the **Permissions** menu in the chat box.

| Mode | What it does |
| --- | --- |
| **Ask first** | Asks before running commands or changing files |
| **Auto** | Follows your permission rules and asks only where they say to |
| **Full access** | Works without asking, with the exceptions below. It goes back to Auto when the app restarts |

In **Ask first**, every action that changes something waits for your approval: shell commands, file edits, code runs, package installs, MCP and connected-app actions, schedules, worktrees and workflows. Edit approvals show the diff before you allow the change. **Auto** and **Full access** keep working on their own after the first reply until the task is done or something needs you.

> **Important:** Full access lets the agent install packages and work outside the project without asking. Use it only in projects you trust, and follow the steps as it works. Permissions guide tool access; they don't replace a sandbox on your machine.

### What always asks first

Some actions ask in every mode, including Full access:

- Pushing code
- Deleting folders or force-deleting files
- Deleting a file through an edit or patch
- Spending money
- Reading secret files
- Changing the folders that define agents, modes and commands

Changes to those folders ask however the path is written, through a link, or by a copy or move into them. Scheduled runs refuse all of these actions and record why. With a Claude plan, Claude Code runs as a full agent, but its tool use still goes through the app's permissions.

### Secret files

Files such as `.env` can hold passwords or keys, so they always ask before they are read, in every mode. This includes reads inside shell commands such as `cat`, `grep`, `head`, `cp` or `source`. The approval names the exact file and says it can hold passwords or keys. If you select **Always allow** on a secret file, it covers only that file, not every secret file.

### Workspace trust

Workspace trust decides which projects' own configuration the app honors. In a folder you haven't trusted, the project's own configuration isn't applied, browser control and device control are ignored, and an external coding agent doesn't load that repository's agent settings. To decide which folders are trusted, run `koder trust`.

### Check your permission policy

Your permissions come from several layers: your access mode, your rules and the project's configuration. These commands show what is in effect:

| Command | What it shows |
| --- | --- |
| `koder inspect` | The effective permission policy and which layer set each value |
| `koder trust` | Which directories may contribute configuration |
| `koder compliance` | Every network destination the configuration allows |
| `koder audit` | A tamper-evident record of permission decisions |

### Set autonomy in the terminal

The terminal app has five levels. Switch with `/auto` followed by the level, for example `/auto ask`.

| Level | What it does |
| --- | --- |
| `ask` | Commands, file changes and network calls wait for your approval |
| `rules` | Follows your permission rules only |
| `read-only` | Reads without making changes |
| `auto` | Works on its own and refuses what is outside Auto, such as pushing |
| `full` | Works without asking, except for the actions that always ask |

When Auto refuses something, such as a push, the terminal says what was outside Auto and how to allow it: switch to Full with `/auto full`, or to Rules or Ask.

## Work with Omni on your desktop

Select **Omni** to hand off work that isn't code: drafting, planning, documents, numbers and the apps you connect in **Omni settings → Connect your apps**. Anything that sends, creates, changes or deletes something is approved one call at a time. Omni in the desktop app and Omni at **[app.aixplain.com](https://app.aixplain.com)** are separate and don't sync yet. See **[Get started with Omni](https://www.aixplain.com/support/articles/get-started-with-omni)**.


## Code with Koder

Select **Koder**, open a project folder and describe a task in plain language, or pick a starter such as find and fix a bug or review my changes. Follow the steps, answer its questions, then read everything the session changed in the **Review** panel and commit from the app. For planning, `AGENTS.md`, MCP servers, skills and CI, see **[Get started with Koder](https://www.aixplain.com/support/articles/get-started-with-koder)**.

## Find security issues with Scan

Scan reads a repository for security bugs and lists what it finds, worst first, each with a trace of how an attacker's input reaches it. Open a project in the sidebar and select **Security scan**, or start a new chat and choose **Check for security bugs**. Scan reads the code but doesn't run it yet, so every finding is a likely issue. See **[Find security issues with Scan](https://www.aixplain.com/support/articles/find-security-issues-with-scan)**.

## Make screens, slides and video with Design

Select **Design** and describe a landing page, a dashboard, an app screen, a slide deck, an animated intro or a vertical video. Design builds it as files in your project and shows it beside the conversation. See **[Make screens, slides and video with Design](https://www.aixplain.com/support/articles/make-screens-slides-and-video-with-design)**.

## Build and deploy agents

Use **Agents** to build an agent with Omni or in code with Koder and the Agents SDK, test it, and deploy it on cloud, where it runs 24/7, or on your own servers. See **[Build and deploy agents](https://www.aixplain.com/support/articles/build-and-deploy-agents)**.

## Use aixplain from the terminal

The same agent, permissions and sessions run in your terminal. Install the CLI:

```sh
curl -fsSL https://github.com/aixplain/koder/releases/latest/download/install | bash
```

On Windows, in PowerShell:

```powershell
irm https://github.com/aixplain/koder/releases/latest/download/install.ps1 | iex
```

Open a new terminal, then run `koder` in a project to open the terminal app, `koder login` to sign in, `koder omni` for Omni, or `koder scan` for Scan. Run `koder --help` for every command in your installed version.

## Pair your phone or tablet

Remote access pairs a phone or tablet with the computer where the app runs, from version 0.5.0 of the app and the `koder` CLI. A paired device can follow chats, answer questions, approve steps and notify you when a chat needs you, all in an installable web app.

1. On your computer, open a terminal and run `koder pair`.
2. Follow what it shows to connect your phone or tablet.
3. On your phone, open the web app and install it if you want it on your home screen.
4. Allow notifications so you know when a chat needs you.

The device connects directly when it is on the same network as your computer, or through an encrypted relay when you are away. Both connections are encrypted, and a device sees only what its pairing allows. The agent keeps running on your computer, so keep your computer on while you follow along.

> **Important:** A paired device can approve steps, so pair only devices you own and keep them locked. A paired device can never raise a chat to Full access.

## Update aixplain Desktop

The app checks for updates when it starts and downloads them in the background. When an update is ready, select **Restart and install**. On macOS, the install step runs before the restart, so the restart is quick. To update by hand, download the current installer from **[aixplain.com/koder](https://www.aixplain.com/koder#download)** and install it over your current app.

Updating keeps your sessions, saved permissions, schedules and theme. On Windows, a failed install can roll back to the version you had. Read the **[release notes](https://www.aixplain.com/release-notes)** for what changed in each version.

### Update the CLI

```sh
koder upgrade
koder --version
```

In the terminal app, a one-time notice and a badge on the status line tell you when a new version is out. Run `/update` to upgrade from there.

> **Important:** Versions 0.2.4 and 0.3.0 can't upgrade themselves. If you are on one of them, run the installer once, as in **[Use aixplain from the terminal](#use-aixplain-from-the-terminal)**. Your configuration, credentials and session history are kept.

`aixplain-code` is the CLI's legacy name. It still works as an alias for `koder`, so scripts that use it keep running.

## Where your data lives

| What | Where |
| --- | --- |
| Global configuration | `~/.config/koder/koder.json` or `koder.jsonc` |
| Sessions and other local state | `~/.local/share/koder` |
| The CLI | `~/.koder/bin` |
| Project instructions | `AGENTS.md` in your repository |

XDG environment settings can change the first two paths. Protect these folders as you would your project files.

## Uninstall aixplain Desktop

- **macOS:** quit **aixplain**, then move it from your Applications folder to the Trash.
- **Windows:** open the Windows settings for installed apps, find **aixplain** and uninstall it.
- **Linux:** remove the DEB or RPM package with your package manager, or delete the AppImage file.
- **The CLI:** run `koder uninstall`. It removes the whole CLI install, including command aliases, the `PATH` entry and other installed versions, with a guard against touching anything outside its own directory.

To remove your local data too, delete `~/.config/koder` and `~/.local/share/koder`, or the paths your XDG settings point to.

> **Important:** Deleting these folders removes your sessions, saved permissions, schedules and settings for good. Keep a copy if you might want them back.

## Pricing and usage

aixplain Desktop is free to download. What you pay depends on the model:

- **aixplain models** draw on your team's credits. One credit is one US dollar, and each model's price is shown in the picker.
- **A connected Claude, ChatGPT or Cursor plan** counts against that plan's limits, not your aixplain credits.
- **Your own provider** bills you directly.
- **Local models** cost nothing beyond your hardware.

To see credits, top up or add a card, open **Settings → Account & usage**. See **[Account, plans and billing](https://www.aixplain.com/support/articles/account-plans-and-billing)** and **[Pricing](https://www.aixplain.com/pricing)**.

## Privacy and data

- **Stored on your computer:** your projects, sessions and settings stay on your machine.
- **Sent to the model you choose:** the parts of your code a task needs go only to the selected model, and hosted providers receive the inputs sent to them. Pick a local model or a zero data retention model when that matters.
- **Crash reports are off by default:** turn them on in **Settings → General → Privacy**. A report says what kind of error happened and where in the app's code, never your messages, files, folder names or keys.

We never sell your data or train on it. See **[Privacy and data](https://www.aixplain.com/support/articles/privacy-and-data)** and **[Security](https://www.aixplain.com/security)**.

## Limitations

- Scan doesn't run your code yet, so its findings are likely issues, not proven ones.
- Omni on the desktop and Omni on the web don't share sessions or settings yet.
- Sign-in, hosted models and connected apps need a network connection.
- Full access goes back to Auto every time the app restarts, and a paired device can't raise a chat to Full access.
- Scheduled runs can't push, delete folders, spend money or read secrets, even when you would approve it.
- The desktop app isn't available for Linux on Arm yet. The CLI and terminal app are.

## Frequently asked questions

### Is it called aixplain Desktop or Koder?

The app installs as **aixplain**. Koder is its coding mode, next to Design and Omni, and the `koder` command runs the same agent in your terminal.

### Does it work offline?

Your sessions and settings are always stored on your computer, and with a local model the model runs there too. Sign-in, hosted models and connected apps need a network connection.

### Which languages does the app support?

The desktop app is available in 17 languages, including Arabic with a right-to-left layout.

### Is Omni on my phone the same as Omni in the app?

No. Omni through WhatsApp or Telegram is the cloud Omni, which is separate from Omni in the desktop app. A paired phone is a remote for the app on your computer. See **[Get started with Omni](https://www.aixplain.com/support/articles/get-started-with-omni)**.

### Can it run in CI?

Yes. `koder run` takes a prompt from the shell, and `koder scan --fail-on high` fails a job when Scan finds something at or above that severity. See **[Get started with Koder](https://www.aixplain.com/support/articles/get-started-with-koder)**.

### Can I add MCP servers and skills?

Yes. Add MCP servers, skills and plugins, and give each project its instructions in `AGENTS.md`. See **[Get started with Koder](https://www.aixplain.com/support/articles/get-started-with-koder)**.

### Who decides which models and connections my team can use?

Your workspace admins. See **[Manage your workspace](https://www.aixplain.com/support/articles/manage-your-workspace)**.

## Troubleshooting

For sign-in, model, connection, update and Omni problems, see **[Troubleshooting aixplain Desktop, Koder and Omni](https://www.aixplain.com/support/articles/aixplain-desktop-troubleshooting)**.

## Get help

- **Report a bug:** use the bug button in the app, or open an issue on **[GitHub](https://github.com/aixplain/koder/issues/new/choose)** with your version, operating system, the steps to reproduce it and what you expected. `koder report` saves a bug report with the session, recent logs and version.
- **See what changed:** read the **[release notes](https://www.aixplain.com/release-notes)** for new features, fixes and known limitations.
- **Report a security issue:** email **care@aixplain.com**. Remove credentials and private data from logs and screenshots first.