Hermes Claude Bridge

hermes-claude-bridge

Use your Claude subscription with Hermes, within the Claude subscription TOS (as of 2026-09-03), via the Agent SDK and `claude -p`, to give an OpenAI-compatible endpoint which harnesses such as Hermes Agent can use.

**READ THE CLAUDE TOS BEFORE USING.** Do NOT hammer it, do not schedule heartbeats every 15 minutes. Don't spread it across multiple users. Use it for modest individual purposes only. Please don't ruin it.

Single user. No legal assurances. Not a recommendation. This is just me sharing my code. Probably don't use this in business contexts. Don't abuse it. It's a thin layer over a recent Agent SDK / Claude Code feature: print mode with structured output.

Policy pages to read first, and re-read later, because the rules changed three times in 2026:

• https://support.claude.com/en/articles/15036540-use-the-claude-agent-sdk-with-your-claude-plan
• https://code.claude.com/docs/en/legal-and-compliance
• https://support.claude.com/en/articles/11145838-use-claude-code-with-your-pro-or-max-plan

Built for [Hermes Agent](https://github.com/NousResearch/hermes-agent). The bridge itself is a plain OpenAI-compatible server (chat completions, streaming, tools, `response_format`), so other harnesses may work. I only tested Hermes. It works so far. Just a side project: Fable 5.1 and I coded it in about two hours. Godspeed o7

Tip: point your personal AI agent at this repo and have it set it up for you.

What it is

Path	Role
claude_bridge.py	OpenAI-compatible server on 127.0.0.1:8790. Every chat completion spawns the unmodified claude binary in print mode with Claude Code's own tools disabled. Stdlib only, Python 3.9+.
plugin/claude-cli/	Hermes Agent model-provider plugin (claude-cli). Forwards /model and /effort.
launchd/	macOS LaunchAgent template so the bridge is always up.
tests/	18 tests against a fake claude script. No usage spent.

Claude is the brain only. Hermes owns th