For AI agents: The documentation index is at https://docs.digitalocean.com/llms.txt. Markdown versions of pages use the same URL with index.html.md in place of the HTML page (for example, append index.html.md to the directory path instead of opening the HTML document).

What is Grok Build (Beta)?

Grok Build is a command-line coding agent (with an interactive terminal UI) from xAI. Describe what you want in natural language and Grok plans the work, edits files, runs commands, and shows clean diffs for review. This image wires Grok Build to DigitalOcean’s OpenAI-compatible serverless inference endpoint, so a single model access key unlocks GPT-5.5, Claude, Llama, Kimi, GLM, DeepSeek, Qwen, MiniMax and more — plus DigitalOcean’s Intelligent Inference Router.

• Terminal-first – Works natively in your shell; no web interface required
• DigitalOcean Serverless Inference – Pre-configured inference with one model access key
• Intelligent Inference Router – Optionally route each prompt to the best-fit model
• Plan mode – Generate a plan, comment on or rewrite steps, then approve before changes
• Parallel subagents – Delegate large tasks to subagents that run in parallel
• Headless mode – Run agents non-interactively in scripts with grok -p
• Works with your conventions – Picks up AGENTS.md, plugins, hooks, skills, and MCP servers

Key Features

• Natural-language, agentic coding assistance in the terminal
• Pre-configured with DigitalOcean Serverless Inference (GPT, Claude, Llama, DeepSeek, and more)
• Optional DigitalOcean Intelligent Inference Router (router:<name>)
• Plan / review / approve workflow with clean diffs
• Headless scripting (-p) with plain, json, and streaming-json output
• xAI account sign-in supported as an alternative provider

System Requirements

Grok Build is lightweight; most compute happens at the inference service.

Included System Components

• Ubuntu 24.04 LTS – Base operating system
• Grok Build – xAI terminal coding agent
• DigitalOcean Serverless Inference – Pre-configured inference provider
• Git – Version control
• curl, jq, unzip – Utilities
• UFW Firewall – SSH only (rate-limited)

Software Included

This Marketplace listing does not include a detailed software list.

Creating an App using the Control Panel

Click the Deploy to DigitalOcean button to create a Droplet based on this 1-Click App. If you aren’t logged in, this link will prompt you to log in with your DigitalOcean account.

Creating an App using the API

In addition to creating a Droplet from the Grok Build 1-Click App using the control panel, you can also use the DigitalOcean API. As an example, to create a 4GB Grok Build Droplet in the SFO2 region, you can use the following curl command. You need to either save your API access token to an environment variable or substitute it in the command below.

curl -X POST -H 'Content-Type: application/json' \
         -H 'Authorization: Bearer '$TOKEN'' -d \
        '{"name":"choose_a_name","region":"sfo2","size":"s-2vcpu-4gb","image":"grokbuild"}' \
        "https://api.digitalocean.com/v2/droplets"

Getting Started After Deploying Grok Build

Getting Started

1. Deploy the Droplet

1. Select this 1-Click App from the DigitalOcean Marketplace
2. Choose a Droplet size (1 GB RAM minimum)
3. Add your SSH key for secure access
4. Create the Droplet

Optional: pass MODEL_ACCESS_KEY (and optionally DO_INFERENCE_MODEL or DO_INFERENCE_ROUTER) as droplet environment variables to skip the wizard.

2. SSH In

ssh root@your-droplet-ip

3. Complete the Setup Wizard

If you passed MODEL_ACCESS_KEY at create time, Grok Build is already configured — skip to step 4.

Otherwise, on first login the setup wizard prompts for your DigitalOcean model access key. To create one:

1. Go to https://cloud.digitalocean.com/model-studio/manage-keys
2. Or from the cloud console, navigate to Inference > Manage
3. Click Create Model Access Key and copy it
4. Paste it when prompted — the wizard then lists current chat models from GET https://inference.do-ai.run/v1/models
5. Pick a default from that live list, or R for the Intelligent Inference Router

The wizard verifies the key against the inference API, rebuilds the DigitalOcean entries in /root/.grok/config.toml from that live list, and sets the default. If the key is rejected you can re-enter it, or type a model id to continue without a live list. Change the model later with /model in the TUI or -m <alias>.

Prefer to use xAI directly? Press Enter at the model-access-key prompt to choose xAI account sign-in or to enter an xAI API key.

No browser needed. With a model access key (or any API key), Grok authenticates with the API directly — there is no OAuth browser step. Because the droplet has no desktop browser, the xAI account path uses device-code sign-in (/opt/grok-login.sh or grok login --device-auth), which shows a short URL and code to open on your laptop or phone. Avoid running a bare grok login, which would try to launch a local browser.

If grok reports no API key in the same session right after setup (or in a non-login shell), load it with source /etc/profile.d/grok-build-key.sh.

4. Run Grok Build

cd /path/to/your/project
grok

For automation:

grok -p "Explain this codebase"
grok -p "Review this diff" --output-format json --always-approve

The default model is the one you picked in the wizard. If you auto-configured without DO_INFERENCE_MODEL, apply prefers GPT-5.5 (openai-gpt-5.5 / alias gpt-5-5) when that id is still in the live catalog; otherwise it uses the first chat-capable model returned by the API. Switch with /model in the TUI or -m <alias> headlessly.

Models

The image does not pin a marketplace catalog. After you enter a model access key, the setup wizard (and /opt/apply-inference-from-env.sh) load chat-capable models from Retrieve available models and rewrite the DigitalOcean [model.*] block in /root/.grok/config.toml. New models appear and deprecated ones drop off. Existing aliases (for example gpt-5-5) are kept when that id is still available. Pick R for the Intelligent Inference Router.

List the live set on the droplet with:

curl -s -H "Authorization: Bearer $MODEL_ACCESS_KEY" \
  https://inference.do-ai.run/v1/models | jq -r '.data[].id'

Use the alias from config.toml with -m or /model. The router alias is router (router:<name>). Re-run /opt/apply-inference-from-env.sh or the setup wizard after new models ship. To add a provider that is not on Serverless Inference, uncomment a [model.<alias>] example at the bottom of /root/.grok/config.toml.

Intelligent Inference Router

DigitalOcean’s Inference Router classifies each prompt and sends it to the best-fit model based on rules you define (optimizing for cost or latency).

1. Create a router under Inference > Routers in the control panel (or via the API) and attach it to the same model access key you use for the droplet.
2. Set DO_INFERENCE_ROUTER=<router-name> in /opt/grok-build.env (or as a droplet env var) and run /opt/apply-inference-from-env.sh, or enter the router name in the setup wizard.

This points the router alias at router:<router-name> on https://inference.do-ai.run/v1 and makes it the default. The router authenticates with the same MODEL_ACCESS_KEY as the direct models — no extra key needed. Using the router is a drop-in replacement for a specific model.

Managing Grok Build

Helper Scripts

Configuration

• Grok config: /root/.grok/config.toml
• API key env template: /opt/grok-build.env (MODEL_ACCESS_KEY, DO_INFERENCE_MODEL, DO_INFERENCE_ROUTER, XAI_API_KEY)
• Active key: /etc/profile.d/grok-build-key.sh (MODEL_ACCESS_KEY or XAI_API_KEY)
• Getting started guide: cat /root/grok_build_info.txt

Start, Stop, Restart, and Update

Grok Build is an interactive CLI, not a long-running background service, so there is no daemon to start, stop, or restart.

• Start a session: grok (or grok -p "..." for headless)
• Stop a session: exit the TUI (Ctrl-C / /exit) or end the headless command
• Restart: simply run grok again
• Update: /opt/update-grok-build.sh (re-runs the official installer for the latest stable release)

Troubleshooting

grok not found in PATH

Ensure you’re in a login shell (SSH session) where PATH is set, or run directly:

/root/.grok/bin/grok

AI features not working

Re-run the setup wizard to reconfigure your DigitalOcean model access key:

/opt/setup-grok-build.sh

Or set the key manually and reload your shell:

echo 'export MODEL_ACCESS_KEY="<your key>"' > /etc/profile.d/grok-build-key.sh
chmod 600 /etc/profile.d/grok-build-key.sh

Then source /etc/profile.d/grok-build-key.sh (or start a new SSH session). Create keys at https://cloud.digitalocean.com/model-studio/manage-keys (Inference > Manage).

Additional Resources

• Grok Build docs: https://docs.x.ai/build/overview
• Custom models reference: https://docs.x.ai/build/overview#custom-models
• Retrieve available models: https://docs.digitalocean.com/products/inference/how-to/retrieve-available-models/
• DigitalOcean Inference Router: https://docs.digitalocean.com/products/inference/how-to/use-inference-router/
• Announcement: https://x.ai/news/grok-build-cli

Support

For Grok Build-specific issues:

• Documentation: https://docs.x.ai/build/overview
• Send feedback from inside the CLI with /feedback

For DigitalOcean Droplet issues:

• DigitalOcean Support: https://www.digitalocean.com/support
• Community Tutorials: https://www.digitalocean.com/community

Note: This 1-Click installs Grok Build via the official xAI installer and pre-configures DigitalOcean Serverless Inference as the inference provider. SSH is the only exposed port; there is no web interface. You can alternatively authenticate with an xAI account (SuperGrok / X Premium+) or an xAI API key.