NEWS · CODING · #443
Grok Build adds background 'workflows' that run tasks across hundreds of parallel agents
Grok Build can now author and run workflows from plain-language requests: it plans multi-phase scripts, fans tasks out to parallel agents (default budget 128 agents, up to 1,024 for large runs), verifies results, saves progress, and reports back when finished. Saved workflows live in .grok/workflows/ or ~/.grok/workflows/, become slash commands, and built-in examples like /deep-research are provided for common tasks such as PR reviews, issue triage, and audits.
KEY POINTS
- Grok Build can now author and run workflows from plain-language requests: it plans multi-phase scripts, fans tasks out to parallel agents (default budget 128 agents, up to 1,024 for large runs), verifies results, saves progress, and reports back when finished.
- Saved workflows live in .grok/workflows/ or ~/.grok/workflows/, become slash commands, and built-in examples like /deep-research are provided for common tasks such as PR reviews, issue triage, and audits.
- This adds a built-in way to scale developer and research tasks into many independent AI agents with saved, repeatable workflows and verification, enabling large audits, PR reviews, and triage to run autonomously in the background.
WHY IT MATTERS
This adds a built-in way to scale developer and research tasks into many independent AI agents with saved, repeatable workflows and verification, enabling large audits, PR reviews, and triage to run autonomously in the background.