Get started

Install Dokima. Run your first review.

We recommend running Dokima on a dedicated Linux VPS, VM or user account so reviews can continue while your team works. You need your Dokima download and activation details, access to your code and Claude, Codex or OpenCode. macOS is also supported.

New to Dokima? Activate Free or start a team trial or buy a subscription. Missing your email? Recover access.

Download and install

Open the download link in your Dokima email (also available in your team account) and choose Linux x86-64, Linux ARM64, macOS Intel or macOS Apple silicon.

Linux

Save the archive to ~/Downloads, or download it directly in your terminal by copying the Linux download link:

mkdir -p ~/Downloads
cd ~/Downloads
wget 'COPIED_LINUX_DOWNLOAD_LINK'

Verify the SHA-256 checksum from the download page and extract the binary (use dokima_linux_arm64.tar.gz on ARM64):

cd ~/Downloads
sha256sum dokima_linux_amd64.tar.gz
tar -xzf dokima_linux_amd64.tar.gz
./dokima version

macOS

In ~/Downloads, verify the SHA-256 checksum and extract the binary (use dokima_darwin_amd64.tar.gz on Intel):

cd ~/Downloads
shasum -a 256 dokima_darwin_arm64.tar.gz
tar -xzf dokima_darwin_arm64.tar.gz
./dokima version

macOS builds are currently unsigned. If macOS blocks the verified binary, allow it in System Settings → Privacy & Security.

Use ./dokima (or ~/Downloads/dokima) in this first terminal session. On first run, Dokima installs the dokima command for future logins without administrator privileges. You can later manage it with dokima update, dokima install and dokima uninstall.

Activate your licence

./dokima activate ACTIVATION_CODE

Use the activation code from your Dokima email or team account. Dokima saves the licence locally. Free provides continuing use for two contributors, including private code; the separate 21-day team trial starts on first activation.

Choose your AI runner and model

Install and configure the command-line tool Dokima will use to analyse your code:

Codex: Install and sign in to Codex CLI, then verify codex --version.
Claude: Install and sign in to Claude Code, then verify claude --version.
OpenCode (Gemini, open models and local models): Install the OpenCode CLI (opencode --version), configure your provider and set a default model. For a local model, point an OpenCode provider at an OpenAI-compatible endpoint that supports streaming Chat Completions and tool calls.

Dokima runs runners in read-only mode and automatically enables the first installed runner it finds. If multiple runners are installed, enable your preferred runner (such as opencode) in Settings and disable the others. Model usage costs and where code is processed depend on your chosen provider or local endpoint.

Open your workspace

cd /path/to/your/workspace
~/Downloads/dokima

Run Dokima in the directory containing the Git repositories you want to review (or run dokima after a fresh login). On first run, approve the workspace and confirm your runner. Dokima stores state in .dokima/ and creates DOKIMA.md for architecture notes and security assumptions. Read more about workspace context.

If you prefer not to give your model Git access, you can have a cron job update the source from Git and copy it to a restricted user account that Dokima runs in.

Run your first review

Choose Full Sweep in the dashboard, or run:

~/Downloads/dokima sweep --mode full

When the sweep finishes, view findings in the dashboard or open ./dokima-report/index.html in your browser. For subsequent reviews, run dokima sweep or see the command guide for continuous mode.

Choose where your team reads the results

Publish results privately to Dokima.net, or, when running on your own infrastructure, serve them on a protected internal company website. The machine running Dokima needs access to the repositories and your chosen AI runner; the report website does not need the repository checkout.

On first launch, choose whether to publish reports privately to your Dokima.net team account or keep Local HTML reports only. Uploading starts only after you confirm and link an organisation, and both options keep the local ./dokima-report/ folder. You can change this anytime in Settings. Learn about team reports. Coding agents can follow the agent setup guide and read hosted findings through the agent API.

Get support Browse documentation