harness and oh-my-harness
These are complements: A provides the agent runtime environment while B generates the guardrails and configuration files (CLAUDE.md, hooks, settings) that constrain and configure agents running within that environment.
About harness
mastersof-ai/harness
Agent runtime with full system prompt control. Define agents in markdown, run them in a terminal TUI or multi-user web UI. Powered by the Claude Agent SDK.
Provides full markdown-based agent definitions with transparent system prompts (no hidden framework instructions), paired with Claude Agent SDK integration and MCP-based tools. Supports both local TUI and multi-user web deployment from the same runtime, featuring persistent memory, sub-agent delegation, and optional sandbox execution—all configurable via standard YAML.
About oh-my-harness
kyu1204/oh-my-harness
Tame your AI coding agents with natural language. Generate enforced guardrails (CLAUDE.md, hooks, settings) from a single command.
Generates project-specific guardrails via Claude AI prompt engineering and an auto-detecting project analyzer covering 14 languages and frameworks. Enforces rules through git hooks, file-system guards, and command-blocking shells with state tracking in JSONL event logs for audit trails. Uses `harness.yaml` as a single source of truth that regenerates `CLAUDE.md`, hook scripts, and `settings.json`—integrating with Claude Code, Cursor, and the standard git pre-commit ecosystem.
Related comparisons
Scores updated daily from GitHub, PyPI, and npm data. How scores work