HarnessMatch
Menu

Coding agent

Grok Build

Extensible Rust coding agent with plan review, sandbox profiles, and ACP.

Status
Active
License
Apache-2.0 first-party code
Evidence
Documented + code-verifiable, 24 sources
Product record checked
2026-07-27
Compare harnessesInspect sources

At a glance

SpaceXAI's open-source terminal coding agent with TUI, headless and ACP surfaces, custom models, MCP, plugins, default-on subagents, plan review, rewind, permissions, and optional OS-level sandbox profiles.

Good choice if

  • Plan-review-execute workflows with explicit diffs and permission controls
  • Terminal, CI, or ACP integrations that need structured output and resumable sessions
  • Extensible setups using MCP, plugins, subagents, custom models, and worktrees

Check before choosing

  • The OS sandbox is off by default; strict/read-only child-network blocking is Linux-only, in-process model and web traffic remains outside that rule, and sensitive paths such as ~/.ssh require explicit deny configuration
  • Worktrees isolate checkouts rather than processes, while always-approve can skip tool prompts unless enterprise requirements explicitly disable bypass mode
  • Cross-session memory is optional and off by default; rewind restores recorded session/file state but does not undo remote services, browser actions, already-started processes, or other external side effects
See 3 more considerations
  • Browser review is supplied through plugins or MCP rather than a core browser tool, and web fetch is separately gated
  • The product remains beta with a rapid 0.2.x cadence; the public repository is a periodic monorepo sync, has no public CI workflow, and does not accept external contributions
  • The statement that Grok 4.5 powers the product is model branding, not evidence of harness quality; no complete independent product benchmark is imported

Why it qualifies as a coding harness

This confirms category fit, not product quality. Every required criterion links back to first-party evidence.

Qualifies4 of 4 required criteria evidenced

  • Adaptive agent loop

    Documented

    The system repeatedly observes results and chooses the next action instead of following a fixed one-pass graph.

    EvidenceGrok Build overview

  • Repository tool execution

    Documented

    The system can use tools to inspect and change a repository or its execution environment.

    EvidenceCLI reference

  • Task-aware context management

    Documented

    The runtime assembles, updates, compacts, retrieves, or persists task-relevant context while work proceeds.

    EvidenceModes and commands

  • Model-independent runtime control

    Documented

    Permissions, budgets, interruption, policy, or stop controls operate outside the model's own text generation.

    EvidencePermissions

Membership establishes category fit only. It does not score quality, safety, autonomy, model capability, or benchmark performance. · Read the membership rule.

How it works under the hood

Seven mechanisms mapped from first-party records. These labels describe what the harness provides, not how intelligent its model is.

7/7layers documented
  • Execution & isolationSandbox availableDocumented mechanism, not a performance score.
  • Tooling & integrationsExtensible toolsDocumented mechanism, not a performance score.
  • Context & statePersistent stateDocumented mechanism, not a performance score.
  • Lifecycle & recoveryCheckpoint/rewindDocumented mechanism, not a performance score.
  • ObservabilityStructured tracesDocumented mechanism, not a performance score.
  • VerificationTool-assistedDocumented mechanism, not a performance score.
  • Governance & permissionsPolicy controlsDocumented mechanism, not a performance score.

Public code audit

3/5public artifacts present
Security policy
Present at inspected commit
CI workflow
Not found
Automated tests
Present at inspected commit
Evaluation assets
Not found
Contributor documentation
Present at inspected commit

The periodic public monorepo sync contains 438 test-like files, implementation guides, and one memory-benchmark example but no public CI workflow or complete coding-task evaluation suite. External contributions are not accepted, and the sync may not exactly identify the shipped binary source revision.

Inspect commit b41c75a578f98bddbd326ab02cd53618451d97ee, checked 2026-07-27

Measured configurations

No benchmark run passes the full metadata admission policy for this harness yet. Missing data is not scored as zero.

Benchmark policy and all runs

Capability support

Documented first-class product support, checked against the sources below.

External tools (MCP)
DocumentedProduct-supported MCP integrationSource · checked 2026-07-27The source establishes the mechanism, not its quality or availability in every mode.
Local models
DocumentedLocal or self-hosted model pathSource · checked 2026-07-27The source establishes the mechanism, not its quality or availability in every mode.
Agent parallelism
DocumentedDelegated or parallel agent workflowSource · checked 2026-07-27The source establishes the mechanism, not its quality or availability in every mode.
Runs without an open UI
DocumentedNon-interactive or automation surfaceSource · checked 2026-07-27The source establishes the mechanism, not its quality or availability in every mode.
Browser control
Not documentedNo first-class support established by the current recordAbsence of current documentation is not proof that the capability is impossible.
Isolated execution
OptionalOptional OS sandbox profiles; disabled by defaultSource · checked 2026-07-27Network and filesystem coverage varies by platform and selected profile.
Undo file changes
DocumentedProduct-supported file or session rollbackSource · checked 2026-07-27The source establishes the mechanism, not its quality or availability in every mode.

Primary evidence

Each capability claim is tied to a first-party record and a verification date.

Product record checked 2026-07-27
Grok Build overviewTUI, headless mode, ACP, custom models, installation, and authenticationOfficial docsHeadless and scriptingOne-shot execution, JSON event streams, sessions, ACP, update control, and CI integrationOfficial docsPermissions and sandboxOS isolation profiles, default-off behavior, filesystem boundaries, and network limitsOfficial docsPermissionsAsk, auto, always-approve, accept-edits and dont-ask modes plus allow and deny rulesOfficial docsWorktreesParallel checkout creation, base refs, subagent isolation, cleanup, and Git limitationsOfficial docsSubagents and extensionsDefault-on subagents plus skills, plugins, hooks, MCP, and LSP extensionsOfficial docsSubagentsBuilt-in and custom child agents, parallel tasks, permissions, models, and worktree isolationOfficial docsHooksTool and session lifecycle scripts, trust, blocking semantics, plugins, and execution scopeOfficial docs
View 16 additional sources
Product and interfaces3 sources
Agent dashboardMulti-session monitoring, agent status, attention requests, attachment, and controlsOfficial docsGrok Build launchSubscription access, plan approval, diffs, and parallel subagentsOfficial announcementCLI referenceGrok Build commands and common flags for prompts, sessions, headless operation, permissions, sandboxing, models, and updatesOfficial docs
Execution and control1 sources
Plan modePlan-only edit gate, approval transition, persisted plan, and permission-mode interactionOfficial docs
Agents, state and recovery3 sources
Modes and commandsPlan review, session branching, rewind, background tasks, and control modesOfficial docsSession and rewind guideLocal session storage, resume, fork, rewind, compaction checkpoints, and recovery boundariesOfficial repositoryMemory guideOptional cross-session memory, default state, storage, retrieval, and lifecycleOfficial repository
Automation and extensions2 sources
Background tasksRecurring loops, background agents, task list, scheduling, interruption, and persistenceOfficial docsMCP serversProject and user MCP configuration, HTTP and stdio transports, OAuth, server scopes, tool approval, and imported compatibility filesOfficial docs
Enterprise and operations4 sources
Enterprise deploymentManaged policy, sandbox pinning, authentication, network, telemetry, ZDR, and bypass lockOfficial docsSettings referenceModels, permissions, sandbox, MCP, subagents, memory, telemetry, and configuration precedenceOfficial docsMonitoring and OpenTelemetry guidePinned session telemetry, OpenTelemetry configuration, local defaults, exported signals, and operational privacy boundariesOfficial repositorySettings overviewTUI and config.toml settings, user and project scopes, starter configuration, precedence, and links to detailed referencesOfficial docs
Releases and public code audit3 sources
Open-source announcementSource release, extension system, MCP, subagents, and local inference configurationOfficial announcementGrok Build changelogBeta status and current 0.2.111 product release verificationOfficial announcementGrok Build repositoryPinned public source sync, runtime, tests, session architecture, sync policy, and licenseOfficial repository