Hermes Acp Plugin — Security skill for Claude Code

acp-hermes

**Cost & quality X-ray plus deterministic governance for [Hermes Agent](https://github.com/NousResearch/hermes-agent)** — via the [Agentic Control Plane](https://agenticcontrolplane.com).

[](https://pypi.org/project/acp-hermes/) [](LICENSE) [](https://pypi.org/project/acp-hermes/)

Two planes, one plugin:

• Local metering (no account, nothing leaves your machine) — every model call and tool call your Hermes agent makes is metered into a local SQLite DB. acp-hermes report shows spend by model, cache hit rate, what actually fills your context, tool error rates, and cost per task. report --json gives agents their own economics so they can self-optimize.
• Cloud governance (optional, free — acp-hermes login) — every tool call is checked against server-side policy (deny / ask / allow with inline approvals), audited with tenant + session attribution, and visible on a team dashboard across all your machines.

Companion to the [Claude Code ACP plugin](https://github.com/agentic-control-plane/claude-code-acp-plugin). Same backend contract, same dashboard, same policies — wired into Hermes's Python plugin system instead of Claude Code's shell hooks.

Full install guide with dashboard walkthrough: [agenticcontrolplane.com/integrations/hermes](https://agenticcontrolplane.com/integrations/hermes)

Install

pip install acp-hermes
hermes plugins enable acp

That's it — local metering is on. After your next Hermes session:

acp-hermes report

ACP local report — last 7 days (this machine, ~/.acp/hermes-local.db)

MODEL SPEND
  model                    calls   in-tok  out-tok  c