Getting started

This guide builds the current workspace version, enables the ZDR guard, and runs one provider-backed model.

Prerequisites

• Rust and Cargo compatible with the repository toolchain
• Git
• a provider API key for the model you intend to use
• Node.js 20 or newer only if you want to run this documentation site

Build the harness

From the repository root:

cargo build -p xai-grok-pager-bin
cargo run -p xai-grok-pager-bin -- --version

During development, use cargo run -p xai-grok-pager-bin -- anywhere this site shows the installed grok command.

Enable the ZDR guard

Create or update ~/.grok/config.toml:

[features]
enforce_zdr = true
telemetry = false
feedback = false

[telemetry]
trace_upload = false

For an auditable launch, set the corresponding process-level guardrails:

export GROK_ENFORCE_ZDR=1
export GROK_TELEMETRY_ENABLED=0
export GROK_TELEMETRY_TRACE_UPLOAD=0
export GROK_EXTERNAL_OTEL=0

The environment values are useful in shells, containers, and CI because the effective policy does not depend only on a user's config file.

Add a provider

Choose one recipe:

• OpenAI API
• DeepSeek
• OpenCode Go
• Codex authentication status

Each recipe reads its secret from an environment variable. Do not paste an API key into config.toml.

Select and run a model

List the catalog:

grok models

Start an interactive session with a configured alias:

grok -m openai-sol

Or run a deterministic smoke test:

grok \
  -p "Reply with exactly: MODEL_OK" \
  -m openai-sol \
  --max-turns 1 \
  --no-subagents \
  --disable-web-search \
  --no-memory \
  --output-format json

In the TUI, use /model <alias>. From the scrollback pane, Ctrl+M opens the model picker.

Run this documentation site

cd docs
npm install
npm run docs:dev

Use npm run docs:build for a production build and npm run docs:preview to serve that build locally.

Next step

Read The ZDR boundary before moving real repository content through an external provider.