Minute Verdict
Grok Build is xAI’s coding agent for the terminal, and it is sold as part of the X subscription rather than on its own. A month of it costs $8 on Plus, and the free tier gives a small allowance of agent requests to try it before paying.
Best for: a developer who already pays for X and works in a terminal, and who wants a Grok 4.7 agent with plan mode, subagents and MCP support inside the same subscription.
Not for: a team that needs per seat billing, cloud agents and pull request review from day one, or a developer who will never open X and does not want to fund it.
From: $8/mo billed monthly, checked November 13, 2026
What Grok Build does
Grok Build is a coding agent that runs in a terminal. It has three surfaces: an interactive full screen interface, a headless mode that answers a single prompt, and the Agent Client Protocol, which lets another app host the same agent. Installation is one command, curl -fsSL https://x.ai/cli/install.sh | bash, for macOS, Linux and WSL, with a PowerShell script for Windows. On the first launch Grok opens a browser to authenticate, and on a machine without a browser the same session starts from an XAI_API_KEY environment variable.
The work happens around a plan. Plan mode blocks every edit until you approve, and the viewer lets you comment on individual steps or rewrite the plan before a single file changes. Subagents run in parallel, each with its own context window and its own worktree, so a research pass, a build and a review can be in flight together. Skills turn a workflow into a slash command, hooks run scripts on file edits or tool calls, AGENTS.md carries the conventions of a directory, and MCP servers connect Linear, Sentry or Postgres. Git integration stages and pushes, sandboxed execution runs untrusted code, and headless mode scripts the agent inside CI.
Since September 21, 2026 the agent runs on Grok 4.7, the model xAI released that day and now ships as the default in the CLI. A custom model can be declared in ~/.grok/config.toml with a base URL and an environment key, and /model switches between them inside a session. The client itself is open source: the repository is xai-org/grok-build, and the front page carries the cargo commands to build, check, lint and format it.
On this page
What you get for the money
| Plan | Price | What is included | The limit that matters |
|---|---|---|---|
| Free | $0/month | Limited AI usage, limited chat requests, limited agent requests | The agent stops early, so a real repository is out of reach |
| Plus | $8/month | Shared weekly usage pool, Grok Chat, Voice, Imagine and Build, Grok API access, Cursor | No Build allowance figure is published |
| Premium | $30/month | 4x the Plus usage, Grok Bot, latest models, code hosting and AI code reviews, cloud agents and automations | The tier where work leaves your machine |
| Super | $100/month | 15x the Plus usage, priority access at peak times | No new Build feature, only more pool |
| Ultra | $200/month | 50x the Plus usage, earliest access to new products | Largest pool |
| Teams | $30+/seat/month | The seat plus team billing, a private marketplace for rules, skills and plugins, centralized administration | Priced per seat, so a small team pays it several times |
| Enterprise | Custom | Pooled usage, invoice billing, deployment controls | Quote only, no self serve price |
Plus at $8 is the entry ticket, the only step that changes what the agent can do rather than how much of it you get. Twelve months on Plus comes to $96. An annual rate exists but is published in the Cursor documentation rather than on the X pricing page, at $80 a year, which is $16 below twelve monthly payments.
The second way to pay is by the token. A developer who sets XAI_API_KEY and never buys a subscription bills the API directly: grok-4.7 costs $2 per million input tokens and $6 per million output tokens, and a fast variant runs at twice the output speed for twice the price. That route has no pool to hit, and it suits a CI pipeline whose bill should follow the job rather than a seat. The pool is shared across Grok and Cursor, refreshes every week, and every price excludes tax.
Where it falls short
There is no way to buy the agent alone. The cheapest seat is the X subscription at $8, and that subscription also pays for the social network and the chat assistant, a fair bundle only if you open both.
The Build allowance is never quantified. The comparison table marks Grok Build as limited on Free, then lists it without a figure on the paid tiers, and the billing FAQ names Chat, Imagine, Voice and the Bot as products drawing on the pool. A developer cannot size a week of agent work against a multiplier applied to a base nobody publishes, and the first month is the only real measurement.
The pricing page carries monthly rates only. The annual rate is documented by Cursor, at $80 a year for the same Plus seat, and a buyer comparing the two pages will find a discount on one and nothing on the other.
Working outside your machine costs $30. Cloud agents, code hosting, AI code reviews and automations all start at Premium, so a Plus seat buys a local agent and nothing else, fine for a solo developer and thin for a team.
The product is young. Version 1.0 shipped on August 7, 2026 and version 1.0.77 on November 5, 2026, and the changelog between them is a list of terminal, Windows console and permission fixes.
Grok Build alternatives
The yardstick below is the $8 Plus seat, the cheapest month in which the agent is usable.
- Cursor, $8 a month through the same X subscription, with the agent inside an editor rather than a terminal. It is not an either or purchase: the same Plus seat carries both, and the question is which surface you open first. Read the Cursor review.
- GitHub Copilot, $10 a month for Pro, the assistant that lives in an editor window and in the pull requests a team already uses. It is the option when code review has to stay on GitHub: AI coding tools.
- Windsurf, now sold as Devin Desktop at $20 a month for Pro, an agent first editor with a free tier. It fits a developer who wants one window for the code and the agent: read the Windsurf review.
The choice between Cursor and Grok Build is not a purchase decision, since both sit inside the same Plus seat. The real decision is whether the work starts in a terminal or in a window, and whether $30 for cloud agents is worth more than $8 for a local one.
Who it is for: Grok Build
Grok Build suits a developer who lives in a terminal and already pays for X, because the marginal cost of the agent is then zero. It suits a small team that wants its conventions in the repository rather than in a wiki, since AGENTS.md, skills and hooks travel with the code. It also suits anyone who wants to read the agent loop before trusting it, which the open source repository makes possible.
It does not suit a team that needs seats, single sign on and central billing, because the self serve ladder stops at individual tiers, with a Teams offer at $30 per seat and no seat price for the cloud agent. It does not suit a developer who wants to click through a window and see the diff in a panel, since the product assumes a keyboard. And it does not suit a buyer who needs a fixed cost per task before committing, because no Build allowance is published anywhere.
How we checked
This is a research review. We have not installed Grok Build, so there is no hands on score, no timing and no sample repository, and the output quality is untested. Every price here was read on November 13, 2026 on the xAI pricing page, and the install steps, the model list and the release dates come from the product page, the documentation and the changelog, all of which answered that day. A first install will add the behavior of the weekly pool, the token cost of a long session and the quality of the patches, and our how we test page covers how a hands on pass is run. Our editorial policy covers how a page like this is funded.
Timeline
- May 17, 2026: the oldest release in the public changelog, version 0.1.209. Source
- August 7, 2026: version 1.0.0, the first stable line. Source
- September 21, 2026: Grok 4.7 is released and now powers Grok Build. Source
- November 5, 2026: version 1.0.77, the current release. Source
What to do next
Install it on the free tier and run one real task in a repository you know: ask for a plan, read it, then approve and read the diff. If the plan is good enough to approve without editing, pay the $8 for Plus and use it for a month, which is where the agent stops being a demo. Move to $30 the first time the work has to run without your machine, or the first time you want a review on a pull request, and not before.
FAQ
Is Grok Build free? The free tier of the X subscription includes Grok Build with a limited number of agent requests. A full working allowance starts on Plus at $8 a month.
How much does Grok Build cost? There is no separate price. Plus costs $8 a month, Premium $30, Super $100 and Ultra $200, and Build is included from Plus up. An API key route bills grok-4.7 at $2 per million input tokens and $6 per million output tokens.
Does Grok Build run on Windows? Yes. Windows uses a PowerShell installer, and macOS, Linux and WSL use a one line bash install. The changelog carries Windows console fixes in the recent releases.
Is Grok Build open source? Yes. The client is published as xai-org/grok-build on GitHub, with the cargo build, check, lint and format commands at the top of the repository.
Which model does Grok Build use? Grok 4.7 since September 21, 2026. Any other model can be declared in the user configuration file with a base URL and an environment key.
Sources
Sources
- https://x.ai/pricing (checked November 13, 2026)
- https://x.ai/build (checked November 13, 2026)
- https://docs.x.ai/build/overview (checked November 13, 2026)
- https://x.ai/changelog/build (checked November 13, 2026)
- https://x.ai/news/grok-4-7 (checked November 13, 2026)
Update log
November 13, 2026: first publication, research review. Prices read on the xAI pricing page the same day, Grok 4.7 token rates read on the release page.
