Structuring complex multi-step tasks with grok workflows

When to use a workflow

Workflows are built for complex, multi-faceted tasks a single conversation cannot hold. Examples include reviewing every feature in a large pull request, triaging the last 100 issues, or auditing a codebase for a specific class of bug. If the work splits into many independent pieces and should end in one clear report, users can request a workflow.

How it works

Grok plans the task as a small script detailing the phases of work, the agents in each phase, and how their results roll up. Each agent starts with a clean, focused context, and the plan can incorporate checks a single pass cannot manage, such as independent skeptics verifying every finding before it reaches the final report. Runs receive a budget of 128 agents, scaling up to 1,024 for larger jobs. Progress is saved continuously during the run, allowing users to pause and resume without redoing finished work. Users can run /workflows to watch progress live, phase by phase, with per-agent token counts.

Saving and reusing workflows

Grok authors the workflow from a user request, smoke-checks it before launching, and improves it between runs, eliminating the need to write scripts manually. Once a workflow works, it can be saved. Workflows stored in .grok/workflows/ are shared with the team, while those in ~/.grok/workflows/ follow the user everywhere. Each saved workflow becomes its own slash command that accepts arguments. For example, a saved pull review can be executed later simply by typing /pr-review 5137. Additionally, /deep-research ships built in, fanning research questions out to parallel investigators, verifying every claim against its sources, and returning a cited report.