API Pay-as-you-go
$1.00/1M input + $2.00/1M output
grok-build-0.1 with 256k context. Tool invocations $5/1k calls.
AgentTool Agent of the Month
Grok Build
xAI's agentic CLI for professional software engineering — Claude Code-compatible, parallel subagents, large-context.
Grok Build is xAI's agentic coding CLI, launched in early beta on May 14, 2026 as a direct competitor to Anthropic's Claude Code and OpenAI's Codex CLI. Built on the Grok 4.3 beta model with a 16-agent Heavy architecture and a reported 2M token context window, it runs as a fullscreen TUI or headless for scripts and CI/CD. Drop-in compatibility with the Claude Code config ecosystem (AGENTS.md, MCP, plugins, skills), up to 8 concurrent subagents, Plan Mode, ACP support. Currently SuperGrok Heavy only ($300/mo, with a $99/mo intro promo for the first 6 months).
+Pros
- 2M token context window — entire large codebases fit in a single session
- 8 concurrent subagents for true parallel execution on multi-module tasks
- Drop-in compatible with Claude Code config ecosystem (AGENTS.md, MCP, plugins, skills)
- Plan Mode gives structured, reviewable multi-step execution before changes are applied
- Headless mode enables use in scripts, bots, and CI/CD pipelines
- ACP support allows integration as agent backend in other IDEs and tools
- Git worktree integration for clean branch-per-subagent workflows
- Custom model support — works with any OpenAI-compatible endpoint
−Cons
- Early beta only — not yet generally available, expect instability and breaking changes
- Requires SuperGrok Heavy subscription ($300/mo) — no free or lower-priced tier yet
- No native Windows support — requires WSL2
- Very new tool — limited community resources, plugins, and third-party integrations
- 2M context window is reported, not officially confirmed by xAI
- No GUI or web interface — purely CLI/TUI, steeper learning curve for non-terminal users
Pricing
Grok Subscription
Premium+
Consumer subscription via grok.com for chat access.
Grok Build is xAI's answer to Claude Code and OpenAI Codex CLI — an agentic command-line coding tool designed for professional software engineering. Launched in early beta on May 14, 2026, it is built on the Grok 4.3 Heavy model with a 2 million token context window, enabling it to reason over entire large codebases in a single session. The tool ships as a rich fullscreen TUI (terminal UI) with mouse support, and also runs headlessly for integration into scripts, bots, and CI/CD pipelines.
The most distinctive technical feature of Grok Build is its multi-agent architecture: it can spin up and coordinate 8 concurrent subagents in parallel, allowing it to tackle multiple modules or tasks simultaneously — a meaningful throughput advantage for large refactors, test generation, or multi-service codebases. Plan Mode provides structured multi-step task orchestration: the agent proposes a plan for the user to review before executing, maintaining human-in-the-loop oversight for critical operations. The tool also supports Git worktree integration, enabling clean branch-per-subagent workflows.
A deliberate strategic choice: Grok Build is fully drop-in compatible with the Claude Code configuration ecosystem. AGENTS.md files, MCP servers, hooks, plugins, and skills all work out of the box — meaning teams already using Claude Code can migrate or run Grok Build in parallel without reconfiguring their toolchain. Agent Client Protocol (ACP) support allows Grok Build to integrate as an agent backend inside other IDEs and applications. Custom model support lets users point Grok Build at any OpenAI-compatible endpoint, making it vendor-agnostic beyond the default Grok 4.3 Heavy.
Key Features
- Grok 4.3 Heavy model: 2M token context window — full large codebase in a single session
- 8 concurrent subagents: parallel execution for multi-module tasks, refactors, and test generation
- Plan Mode: multi-step orchestration with human review before execution
- Claude Code config compatibility: AGENTS.md, MCP servers, hooks, plugins, skills — zero reconfiguration
- TUI (Terminal UI): rich fullscreen interface with mouse support
- Headless mode: run in scripts, bots, CI/CD pipelines without user interaction
- ACP (Agent Client Protocol): integrate as agent backend in other IDEs and tools
- Custom model support: point at any OpenAI-compatible endpoint
- Git worktree integration: branch-per-subagent workflows
- API key auth: GROK_CODE_XAI_API_KEY for non-browser / non-interactive environments
- Docs: docs.x.ai/build/overview
- Early beta: SuperGrok Heavy subscribers only ($300/mo) — general availability TBD
- Direct competitor to Anthropic Claude Code and OpenAI Codex CLI
Version History
Initial public beta launch. Grok 4.3 Heavy model, 2M token context window, 8 concurrent subagents, Plan Mode, fullscreen TUI with mouse support, headless mode for CI/CD, Claude Code config compatibility (AGENTS.md, MCP, plugins, skills), Git worktree integration, ACP support, custom OpenAI-compatible model endpoints. SuperGrok Heavy subscribers only ($300/mo, $99/mo intro promo).
Signature Snippet
A senior engineer clones a large monorepo and launches Grok Build in Plan Mode: the agent indexes the codebase, proposes a multi-step refactoring plan across 12 files, then executes it with parallel subagents — each working in its own Git worktree on a separate module. All existing MCP servers and AGENTS.md guidelines from the team's Claude Code setup are picked up automatically with zero reconfiguration, and the developer can interrupt with /feedback at any point to course-correct.More in this Space
Prime Agent (Prime Intellect)
Prime Agent is an open-source (MIT) self-improving coding and research harness by Prime Intellect, built on a Recursive Language Model that runs inside a persistent IPython REPL and a Continual Harness that refines its own state via /refine. It took the same frontier weights from 30.2% to 95.5% on ARC-AGI-3 RHAE Best@1, powered daemon-backed sessions, and ships no hosted offering.
OpenCodeReview (ocr)
OpenCodeReview (ocr) is Alibaba's open-source AI code-review CLI. It reads git diffs and drives an LLM agent through a hybrid architecture where deterministic pipelines handle file selection, rule matching, and comment positioning while the agent does the reasoning. Built on two years of internal use at Alibaba scale, it ships a multi-language ruleset and consumes roughly 1/9 of the tokens of general-purpose agents.
Delta (by Zed)
Multiplayer environment from Zed for coding with agents. Threads unify the agent conversation, the code edits, and human review in one shareable artifact, replacing the branch/PR loop. Built on DeltaDB, a CRDT-based version control layer that records every edit between commits alongside the prompts and reasoning that produced them.