</>MCP Agents Market
Plugin

claude-code-harness

by Chachamaru1273.1kShellUpdated 2026-08-31

Claude Code Dedicated Development Harness - Achieving High-Quality Development Through an Autonomous Plan→Work→Review Cycle

Claude CodeCodex CLICursorGrok

Claude Code Harness is a development workflow plugin that enforces a disciplined Plan→Work→Review→Release cycle for AI coding agents. It prevents common agent coding drift by requiring explicit specification approval before implementation, independent review after coding, and evidence-based release packaging. The plugin includes Go-native guardrails that intercept and validate every tool call before execution, blocking risky operations like secret access, network egress, and production deployments. It supports Claude Code, Codex CLI, Cursor, and Grok, providing five core commands that guide agents through structured development stages while maintaining audit trails and task consistency.

Key Features

Five-verb workflow loop: /harness-plan, /harness-work, /harness-work all, /harness-review, /harness-sync, and /harness-release commands
Pre-execution guardrail engine written in Go that validates every tool call before it runs, with 15 tunable rules (R01-R15) plus 5 non-overridable runtime floor categories
Automatic generation of spec.md and Plans.md from intent, requiring human approval before implementation begins
Independent review stage that separates implementation from verification, blocking completion on major findings
Session roster and messaging system allowing multiple AI agents on the same repository to coordinate and share state
Three single-screen HTML decision surfaces (Plan Brief, Progress, Acceptance) for non-engineer stakeholders to judge work without reading code
Machine-checked claims with CI gates verifying component wiring, task ledger consistency, and reproducible builds
Optional Breezing mode for multi-agent (Planner/Critic/Worker) team execution and harness-mem integration for cross-session memory

Use Cases

  • 01Preventing AI coding agents from drifting off-spec during long implementation runs by enforcing plan approval gates
  • 02Ensuring test-driven development compliance by blocking task completion when tests are required but missing
  • 03Coordinating multiple AI agent sessions working on the same repository through shared roster and message passing
  • 04Generating auditable release evidence with CHANGELOG, tags, and verification records linked to acceptance criteria
  • 05Protecting production systems by blocking network calls, secret reads, and deployment operations outside approved task scope
  • 06Enabling non-technical stakeholders to approve plans and accept deliverables through simplified HTML decision surfaces

Related Plugins

View more

andrej-karpathy-skills

A single CLAUDE.md file to improve Claude Code behavior, derived from Andrej Karpathy's observations on LLM coding pitfalls.

Plugin216.1k

claude-mem

Persistent Context Across Sessions for Every Agent – Captures everything your agent does during sessions, compresses it with AI, and injects relevant context back into future sessions. Works with Claude Code, OpenClaw, Codex, Gemini, Hermes, Copilot, OpenCode + More

Plugin95k

Understand-Anything

Graphs that teach > graphs that impress. Turn any code into an interactive knowledge graph you can explore, search, and ask questions about. Works with Claude Code, Codex, Cursor, Copilot, Gemini CLI, and more.

Plugin84.9k

rtk

CLI proxy that reduces LLM token consumption by 60-90% on common dev commands. Single Rust binary, zero dependencies

Plugin82.1k

claude-code-harness — FAQ

What is Claude Code Harness?+

Claude Code Harness is a plugin that adds a structured development workflow to AI coding agents, replacing ad-hoc prompting with a repeatable Plan→Work→Review→Release cycle. It includes a Go-based guardrail engine that validates every command before execution to prevent drift and unsafe operations.

How do I install the Claude Code Harness plugin?+

In Claude Code, run '/plugin marketplace add Chachamaru127/claude-code-harness' followed by '/plugin install claude-code-harness@claude-code-harness-marketplace', then run '/harness-setup'. For Codex CLI, Cursor, or Grok, run the corresponding setup script from the scripts/ directory and restart your tool.

Which AI coding tools work with Claude Code Harness?+

Officially supported tools are Claude Code (v2.1+), Codex CLI, Cursor, and Grok. OpenCode has internal-compatible tier support. Hermes Agent, Codex app, GitHub Copilot CLI, and Antigravity CLI are in candidate or future tiers with varying levels of functionality.

Do I need API keys or external services to use Harness?+

No external API keys are required for the core workflow and guardrail engine. You only need write access to your repository and Claude Code v2.1+ (or another supported tool). The optional harness-mem plugin adds cross-session memory but is not required.

Is Claude Code Harness free and open source?+

Yes, it is released under the MIT license and available as open source on GitHub. All core functionality including the guardrail engine, workflow commands, and safety layers are included at no cost.

What are the guardrails and how do they protect my project?+

Guardrails are two-layer safety checks: runtime floors (non-overridable blocks on billing, network, secrets, production, and deletion) and 15 tunable rules (R01-R15) that deny, confirm, or warn on risky operations. Every tool call is validated by a Go engine before execution, with all decisions logged to JSONL for audit.

How do I install claude-code-harness?+

Open the source repository on GitHub and follow its README. claude-code-harness is a plugin — MCP Agents Market links you directly to the official repo.

Is claude-code-harness free?+

claude-code-harness is an open-source project hosted on GitHub. Check the repository for its license and any usage requirements.

Related searches

Developer Tools PluginsCode Review PluginsWorkflow PluginsBest MCP ServersBest SkillsBest AgentsTop rated listingsClaude Code toolsBrowse all categories