Grok Build moves from one session to many

xAI is pushing Grok Build toward a control surface for parallel coding work, not just a single terminal assistant. In a June 15 tweet, the Grok account wrote: “Manage multiple agents at once in the Agent Dashboard.” The post says users can see what each agent is doing, reply to sessions that need input, and dispatch new tasks from the dashboard.

The product page gives the concrete mechanics. Agent Dashboard gathers every Grok Build session on one screen, sorts sessions by state, and pulls anything waiting for input to the top. xAI’s example shows 3 agents at once: 2 working and 1 idle. Users can launch it from the shell with grok dashboard, or from inside a session with /dashboard or Ctrl+\.

That matters because coding-agent tools are increasingly limited by coordination, not raw code generation. A developer can already ask an agent to edit files or run tests, but managing several concurrent attempts across repositories is harder: approvals get missed, questions arrive in different sessions, and work can continue after the user has lost track of context. xAI says the dashboard groups sessions by working directory, rolls subagents under the parent session that launched them, and lets users peek at recent output without opening the full conversation.

Grok Build itself arrived in early beta in late May for SuperGrok and X Premium Plus subscribers as a terminal-based coding agent. The dashboard adds a scheduling and supervision layer on top of that beta. It also shows where the category is heading: the winning interface may be less about a chat box and more about a compact operations view for many autonomous coding runs.

Watch next for reliability details: how approvals are handled, how conflicts between sessions are surfaced, and whether the dashboard can preserve enough context for developers to trust multiple agents editing real repositories in parallel.

#grok-build

Grok Build dashboard puts parallel coding agents on one screen