QUICK ANSWER
Grok Build is xAI's new coding agent - an interactive TUI running on Grok 4.3 with headless scripting and Agent Client Protocol (ACP) support. Currently in beta, available to SuperGrok ($30/month) and X Premium+ subscribers. Grok Build enters the coding agent market directly against OpenAI Codex and Anthropic Claude Code. Key differentiator: native X/Twitter data access for real-time context. Key gap: no desktop computer use, shorter context window than Claude Code, early-stage tooling vs mature competitors.
What Grok Build Actually Is
Grok Build is xAI's terminal-native coding agent - a command-line interface with an interactive TUI (text-based UI) that provides a richer experience than a plain shell prompt. The design is closer to Claude Code (which also runs in the terminal) than to OpenAI Codex (which is desktop app and cloud-first). You install Grok Build, point it at a project directory, and interact with it through the terminal to plan, write, modify, and run code.
Headless scripting support means Grok Build can be invoked non-interactively - you pass a task as a command-line argument and it executes without requiring a TUI session. This enables Grok Build to run inside CI/CD pipelines, GitHub Actions workflows, cron jobs, and other automated contexts where a human is not present to interact with the TUI. This is the "while-I-sleep" automation use case that has driven developer interest in all three coding agents.
The agent runs on Grok 4.3 - the current xAI flagship model, stable release as of April 17, 2026. Grok 4.3 is the model that introduced Skills - saved instruction sets that let the model remember your preferred code style, project conventions, and workflow patterns without re-explaining them every session. Skills carry into Grok Build sessions, meaning the agent learns and retains your preferences over time.
Agent Client Protocol (ACP) - What It Is and Why It Matters
Grok Build launches with native Agent Client Protocol (ACP) support. ACP is a protocol specification for how AI agents communicate with each other and with orchestration systems - essentially a standard message format and capability declaration mechanism for multi-agent coordination. ACP is distinct from Anthropic's Model Context Protocol (MCP), which is primarily a standard for how AI systems connect to tools and data sources.
| Protocol |
Primary Purpose |
Created By |
Supported By |
| MCP (Model Context Protocol) |
Agent-to-tool and agent-to-data-source connections |
Anthropic |
Claude Code, Cursor, most major AI tools (3,000+ servers) |
| ACP (Agent Client Protocol) |
Agent-to-agent communication and orchestration |
IBM / Linux Foundation (open standard) |
Grok Build, IBM Bee Agent, BeeAI framework |
ACP support means Grok Build can participate in multi-agent orchestration systems as a first-class agent - it can receive tasks from an orchestrator, report status, delegate subtasks to other ACP-compatible agents, and return structured results. In practice, this means Grok Build can be wired into enterprise agent pipelines alongside other ACP-compatible tools. MCP remains the dominant protocol for tool integrations in the AI agent ecosystem; ACP is earlier-stage but has IBM's backing and Linux Foundation governance behind it, which gives it credibility in enterprise contexts.
Grok Build vs Codex vs Claude Code
| Feature |
Grok Build (xAI) |
Codex (OpenAI) |
Claude Code (Anthropic) |
| Interface |
Terminal TUI |
Desktop app + CLI + mobile |
Terminal CLI |
| Headless / CI support |
Yes |
Yes |
Yes |
| Real-time data access |
X/Twitter + web (native) |
Web search (tool) |
Web search (tool) |
| Context window |
256K tokens (Grok 4.3) |
128K (272K configurable) |
1M tokens (beta) |
| Agent protocol |
ACP (IBM/Linux Foundation) |
OpenAI tool spec |
MCP (3,000+ servers) |
| Desktop computer use |
No |
Yes (Windows + Mac) |
Via Cowork |
| Async / fire-and-forget |
Via headless scripting |
Yes (core design) |
Via Managed Agents (enterprise) |
| Pricing entry point |
SuperGrok $30/mo |
ChatGPT Plus $20/mo |
Claude Pro $20/mo |
| Status |
Beta |
GA |
GA |
Grok Build's genuine differentiator is native X/Twitter data access. The Grok 4.3 model has direct access to X's full firehose in real time - not web search that occasionally surfaces X posts, but actual real-time access to the platform's data. For coding tasks where real-time community signals matter - "is this library deprecated?", "what are people reporting about this API behavior?", "did this service have an outage in the last hour?" - Grok Build can answer from live data where Codex and Claude Code can only search the web.
The gaps are real too. Grok Build is beta software entering a market where Codex and Claude Code have both been GA for months. The MCP ecosystem (3,000+ servers) dwarfs ACP's current integration catalog. There is no desktop computer use. The 256K context window is better than Codex's default but far short of Claude Code's 1M beta. For developers already productive with Codex or Claude Code, the case for switching to a beta product is narrow today. The case for watching closely is strong - xAI is using only 11% of available Colossus compute for Grok models per Elon Musk, which means the model improvements from a compute ramp are coming.
Frequently Asked Questions
How do I access Grok Build beta?
Grok Build beta is available to SuperGrok ($30/month) and X Premium+ subscribers. Install instructions are in the Grok Build documentation at docs.x.ai. The CLI installs via the standard package manager for your OS; authentication uses your xAI API key.
What is xAI's Skills feature and how does it work with Grok Build?
Skills, launched with Grok 4.3 across grok.com, iOS, and Android, are saved instruction sets that persist across sessions. You can create a Skill by describing your preferences in conversation, uploading a file, or writing one from scratch. In Grok Build, Skills carry over - if you have a Skill that defines your project's coding conventions, the agent will apply them automatically without you re-specifying them each session.
Is xAI still a separate company from SpaceX?
xAI merged with SpaceX and is being rebranded as SpaceX AI, per Elon Musk's confirmation in May 2026. The Grok product line and xAI developer APIs continue under this structure. The operational impact for developers - API endpoints, model availability, pricing - has not changed. The brand consolidation is happening gradually; xAI as a legal entity is being wound down into SpaceX's corporate structure.
When will Grok Build exit beta?
xAI has not announced a GA timeline. The beta label typically means the product is functional but the interface, features, and pricing may change before general availability. For production use in CI/CD pipelines, treat Grok Build as experimental until xAI confirms GA status and a stable API contract.