Amplifying/agent-intelligence
/coding-agentsgrok-build

xAI

Grok Build

xAI's terminal coding agent, open-sourced under Apache 2.0 on July 15, 2026 (xai-org/grok-build, ~845K lines of Rust): plan mode with reviewable steps, parallel subagents isolated in git worktrees, /goal long-running autonomous mode, headless mode for scripts and CI, and ACP support, with AGENTS.md, plugins, hooks, skills, and MCP servers working out of the box. Can be compiled and run fully local-first against your own inference via config.toml.

CLIBetaOpen sourceBYOKMCP Yes

n/a

Attributed output

No signal we can detect yet

#34

Index rank

of 35 tracked agents

n/a

Business signal

No sourced business metric

$30/mo SuperGrok

Entry price

Plan usage limits, undisclosed

6,033

GitHub stars

xai-org/grok-build

Not published

npm demand/wk

Public npm counter

Market

Developer output

No signal we can detect yet

Grok Build works through the developer's own identity and leaves nothing public we can currently attribute to it on its output, so its public footprint is invisible \u2014 not absent, and not a sign of low usage. Agents in this group are ordered by their sourced business and traction signals instead.

Index rank
#34 of 35
Demand share
No public npm package
Business signal
No sourced business metric
GitHub share
0.5% of tracked stars
npm channel
No public npm package

Pricing And Economics

Expand

Beta CLI included with SuperGrok ($30/mo) and X Premium Plus ($40/mo); headless use bills an xAI API key at published token rates. Since the July 15, 2026 Apache 2.0 open-source release, the harness can also be compiled and run fully locally with your own inference at no charge.

Entry price
$30/mo SuperGrok
Top self-serve price
$300/mo
Pricing unit
Plan usage limits, undisclosed
Token transparency
Retail token equivalent
Included usage
xAI has not published Grok Build usage quotas for SuperGrok, X Premium Plus, or SuperGrok Heavy during the beta. Server-side usage limits were reset for all users alongside the July 15 open-source release.
Overage path
Headless and API paths authenticate with an xAI API key billed per token; no published in-plan overage mechanics yet. The self-compiled local-first path has no cloud caps.

Retail Token Equivalent

No subscription token quota is published. At the reported grok-build-0.1 API rates ($1 per 1M input, $2 per 1M output, ~$0.20 cached input), $30 equals about 30M input or 15M output tokens; this is dollar-equivalent math, not a plan entitlement.

PlanPriceAllowanceUnitToken / Overage Notes
SuperGrok$30/moGrok Build beta access; usage quotas not published.Plan limits, undisclosed$30 at reported grok-build-0.1 API retail = ~30M input or ~15M output tokens.
X Premium Plus$40/moGrok Build beta access bundled with X platform perks; usage quotas not published.Plan limits, undisclosed
SuperGrok Heavy$300/moHighest Grok tier; some launch coverage described the earliest beta as Heavy-only.Plan limits, undisclosed
xAI API keyUsage-basedHeadless/CI and ACP use with an xAI API key, billed per token; new API accounts include $25 in free credits.Per-tokenReported grok-build-0.1 rates: $1 per 1M input tokens, $2 per 1M output tokens, ~$0.20 per 1M cached input tokens.
Self-compiled (open source)$0Apache 2.0 source (xai-org/grok-build) compiled locally and pointed at your own inference via config.toml; no cloud caps.BYO inference

Included with SuperGrok ($30/mo) and X Premium Plus ($40/mo) subscriptions, with SuperGrok Heavy at $300/mo; headless and API use authenticate with an xAI API key billed at metered token rates. Since the July 15 open-source release, the harness can also be compiled from source and run fully locally with your own inference at no charge.

  • The grok-build-0.1 model is reported at roughly $1 per 1M input tokens and $2 per 1M output tokens on the API path, with cached input around $0.20 per 1M; new API accounts include $25 in free credits.
  • Server-side usage limits were reset for all users alongside the July 15, 2026 open-source release; the self-compiled local-first path has no cloud caps.
  • No Grok Build usage quotas are published for any subscription tier during the beta.

Product

Providers

xAI

Model families

GrokCustom models

Capabilities

Terminal agentPlan modeParallel subagentsGit worktreesGoal modeHeadless modeACPMCPPlugin marketplaceSession import (Claude Code/Codex/Cursor)Custom/local models

Tradeoffs

  • Beta
  • Open license, closed governance: CONTRIBUTING.md states external contributions are not accepted; the repo history is a single bulk bot commit with updates pushed from xAI's internal monorepo
  • July 2026 privacy incident: the CLI uploaded entire working directories to xAI's cloud storage regardless of privacy settings; xAI disabled the feature, deleted retained data, and turned retention off by default
  • Leaves no default public attribution, so its output cannot be counted