xAI has introduced `/goal`, a new mode inside its terminal-based coding agent Grok Build that allows developers to set long-running autonomous coding tasks with built-in verification.
The feature works by accepting a single-line objective, for example, `/goal Migrate the auth module to the new API`, after which the agent plans an approach, breaks the work into a checklist, executes items one by one, and verifies the outcome before marking the task complete. Verification can include reviewing generated code, inspecting web pages to confirm behaviour, or running test scripts.
Developers can monitor progress with `/goal status`, pause or resume execution, or add further instructions mid-run. The feature is designed for multi-step, mechanical, testable tasks such as module migration, dependency upgrades, refactoring with test suite validation, or porting configuration across files.
Grok Build already supports plan mode, parallel subagents, AGENTS.md conventions, plugins, hooks, MCP servers, and headless mode for CI pipelines. The `/goal` addition extends the tool’s capability toward unattended autonomous execution, the agent does not stop after writing code but continues until the result is verified.
Access requires a SuperGrok or X Premium Plus subscription. Grok Build can be installed with a single curl command: `curl -fsSL https://x.ai/cli/install.sh | bash`.
The launch positions xAI alongside Anthropic’s Claude Code, OpenAI’s Codex CLI, and Cursor in the competitive AI coding agent space, with `/goal` differentiating on built-in verification as part of the autonomous loop rather than an optional afterthought.
Sources: xAI Launches /goal in Grok Build (MarkTechPost, June 22, 2026); Introducing Grok Build (xAI, May 25, 2026)