MetaHarness adds rules and checks around Hermes-style agents
MetaHarness is an NPM package that sits around existing AI tools such as Claude Code, OpenAI Codex, , and Hermes. It creates a separate for each , with task rules, tool policies, scoring, receipts, routing, verifiers, and .
Each run leaves evidence behind, and each failure becomes a signal for improving the harness. A new improvement is not promoted until it passes tests, cost checks, and safety gates.
Supported execution surfaces include Claude Code, OpenAI Codex, , OpenCode, OpenClaw, Hermes, pi.dev, RVM, and . The package passed 100,000 downloads in its first two weeks.
Key points
- MetaHarness supports Hermes alongside several other .
- It builds repository-specific rules and tool policies for agent runs.
- Every run creates evidence, so failures can guide later improvements.
- Changes must pass tests, cost checks, and safety gates before promotion.
- Hermes Agent users should test it on a small repo before relying on it for larger work.