HarnessMatch
Menu

Coding agent

Command Code

Terminal-first harness with policy controls, subagents, and checkpoints.

Status
Active
License
Proprietary
Evidence
Documented, 21 sources
Product record checked
2026-07-27
Compare harnessesInspect sources

At a glance

A multi-model terminal harness with interactive and headless modes, parallel subagents, MCP, policy-based permissions, worktrees, session checkpoints, hooks, skills, and IDE context integration.

Good choice if

  • Terminal users who need explicit allow, ask, and deny policy rules
  • Parallel exploration and implementation with constrained subagents
  • Headless CI runs that must fail closed instead of waiting for approval

Check before choosing

  • Project path controls and Git worktrees constrain file organization but do not provide an OS security boundary for shell execution
  • Headless editing requires `--yolo`, which enables writes and shell commands together rather than granting a narrow mutation capability
  • Checkpoints are per-session, omit files over 10 MB, and cannot reverse shell or external-service side effects
See 2 more considerations
  • Goal completion is judged inside the product workflow rather than independently replicated evidence; unattended goals still need explicit budget and stop policies
  • Monitors, cron schedules, background commands, and background subagents inherit the host execution boundary unless the user supplies an external sandbox

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.

    EvidenceGoal mode

  • 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.

    EvidenceMemory instructions

  • 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 & isolationWorkspace isolationDocumented 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.
  • ObservabilityLogs/transcriptsDocumented mechanism, not a performance score.
  • VerificationTool-assistedDocumented mechanism, not a performance score.
  • Governance & permissionsPolicy controlsDocumented mechanism, not a performance score.

Public code audit

Unrankedsupport-only repository
Security policy
Not found
CI workflow
Not found
Automated tests
Not found
Evaluation assets
Not found
Contributor documentation
Present at inspected commit

The public tree contains a product readme, documentation links, and issue configuration but no tags, core agent implementation, CI workflow, tests, security policy, or evaluation suite. The shipped npm package therefore cannot be source-audited from this repository.

Inspect commit a774fe8cbe71697d115d4660de299c9c1b286cea, 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
Not documentedNo first-class support established by the current recordAbsence of current documentation is not proof that the capability is impossible.
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
Not documentedNo first-class support established by the current recordAbsence of current documentation is not proof that the capability is impossible.
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
CLI referenceInteractive and print modes, sessions, worktrees, model selection, permissions, and IDE setupOfficial docsCustom agentsParallel subagents, isolated context, tool policy, background execution, and model selectionOfficial docsPermissionsDefault read-only access, approval-gated writes and shell, project trust, local path scope, optional telemetry, and headless safety behaviorOfficial docsMCP integrationMCP server configuration, tool exposure, authentication, and permission integrationOfficial docsHeadless modeRead-only automation default, explicit yolo mutation mode, ten-turn default, exit codes, transcript persistence, and deterministic resume flagsOfficial docsFile checkpointsAutomatic per-message backups, file and conversation restore modes, 10 MB exclusion, per-session scope, and all-or-nothing restore validationOfficial docsHooksPre-tool policy decisions, stop checks, session context injection, audit hooks, retry caps, execution semantics, and failure behaviorOfficial docsAgent skillsProgressively loaded project and user skills, direct invocation, trust boundaries, and reusable workflow instructionsOfficial docs
View 13 additional sources
Product and interfaces1 sources
Interactive modeTerminal session surface, input modes, keyboard controls, editor handoff, model switching, and user interaction defaultsOfficial docs
Execution and control2 sources
Git worktreesSession isolation in Git worktrees through the slash command, CLI flag, and agent-controlled enter and exit toolsOfficial docsPlan modeReasoning-before-execution mode, plan review, inline comments, revision, approval, and auto-accept behaviorOfficial docs
Agents, state and recovery4 sources
Sessions and checkpointsDurable JSONL transcripts, resume, fork, clone, scratchpad, sharing, export, compaction, checkpointing, and rewind behaviorOfficial docsGoal modeStanding multi-turn objectives, autonomous continuation, verification loop, progress handling, and explicit goal cancellationOfficial docsMemory instructionsAGENTS.md instruction tiers, project and user scopes, path imports, precedence, and assembly into agent requestsOfficial docsTaste profilesLearned preference profiles, enable and disable controls, review, sharing, management commands, and persistence boundariesOfficial docs
Automation and extensions2 sources
Background tasks and schedulingBackground commands, monitors, scheduled wakeups, task ledger, background subagents, cron scheduling, sleep, and TUI managementOfficial docsMods extension APITypeScript extension API for tools, commands, lifecycle hooks, event observers, rendering, model providers, packaging, and verificationOfficial docs
Enterprise and operations2 sources
Settings and configurationUser and project JSON configuration, environment variables, MCP and keybindings, scopes, precedence, and documented defaultsOfficial docsCommand Code StudioHosted dashboard, Taste Studio, usage and billing, API keys, organization administration, and account control surfaceOfficial docs
Releases and public code audit2 sources
Public support repository snapshotImmutable public support and product-information snapshot; it does not expose the proprietary agent implementation, tests, or evaluation suiteOfficial repositoryCommand Code v1 changesVersioned summary of the rewritten permission engine, worktrees, sessions, background tasks, context handling, and extension changesOfficial announcement

Ecosystem discovery

OpenRouter coding apps Discovery signal only; usage rank is not used as a quality or capability score. Observed 2026-07-27.