What is harness?

harness is a unified CLI that runs any of four coding agent CLIs — Claude Code, OpenAI Codex, OpenCode, or Cursor — through a single interface.

It spawns the selected agent as a subprocess, translates its native streaming output into a unified NDJSON event stream, and outputs it to stdout. Write one integration, and it works with any supported agent backend.

Why harness?

Every coding agent has its own CLI, its own flags, its own output format. If you're building tooling on top of coding agents — orchestration systems, CI pipelines, evaluation harnesses, or custom UIs — you'd normally need to write a separate integration for each one.

Harness solves this by providing:

• One CLI — harness run --agent <name> --prompt "..." works the same regardless of which agent you pick
• One event format — 7 unified event types (SessionStart, TextDelta, Message, ToolStart, ToolEnd, Result, Error) with consistent JSON schemas
• Model aliases — --model sonnet resolves to the right model ID for whichever agent you're using
• Process safety — subprocess management with process groups, graceful shutdown, and stderr capture

Supported agents

Next steps

• Install harness
• Quick start guide
• Learn about the unified event stream