Rig: A Control Plane for Coding Agents

Описание и применение

Rig обеспечивает контроль над кодирующими агентами, устраняя необходимость ручного проверки и гарантируя, что изменения проходят реальные проверки перед слиянием.

Задача

Developers repeatedly babysit coding agents that claim completion without passing real checks, leading to wasted time and merge failures.

Результат

Rig reduces manual babysitting, ensures verified completion, and streamlines parallel agent execution without nested agent spawning.

Рабочий процесс

• 1. Configure parent and workers in .rig/harness.toml
• 2. Run rig init and rig doctor
• 3. Issue a prompt to the parent CLI (e.g., "Fix the failing tests in tests/test_cli.py")
• 4. Parent scopes files and writes a brief
• 5. Worker(s) execute tasks over MCP
• 6. Parent verifies requirements, runs checks, and accepts if all pass
• 7. Merge changes to main

Доказательства

Rig is not a chat box that "codes for you." It is the harness between agents you already use.

Источник

Parent verifies with real checks. Done means verified, not exit 0 vibes.

Источник