muse-bridge

What muse-bridge does

muse-bridge adds a model named ms/muse to a self-hosted 9Router so that Hermes Agent can select Muse like any API model. Muse has no callable API, so the bridge works through a file queue. Requests from 9Router are written as files, a courier checks the queue about every 5 seconds and wakes Muse through hooks, and Muse's answer is carried back. The model supports native tool calls, so Muse can ask for commands to run on the VPS, and it can understand images.

Dangerous commands still need user approval, and the design avoids opening a new port to the internet. A combo called muse contains ms/muse, so it can be called by that short name. The README is written for complete beginners, with a glossary and tutorials in English and Indonesian. It warns that each round takes roughly 20 to 30 seconds and will not reach the 2 to 5 seconds of a real API.

Key features

• Adds ms/muse as a model in 9Router, usable from Hermes like any other model
• File-queue bridge polled about every 5 seconds
• Native tool calls so Muse can request commands on the VPS
• Image understanding for pictures sent through Hermes
• Beginner tutorials in English and Indonesian

When to use it

• Using a Muse agent from Telegram through Hermes
• Having an agent check disk space, services or logs on a VPS with real command output
• Offering Muse as one model choice next to ChatGPT or DeepSeek in 9Router

Who it is for: Self-hosters who already run 9Router and Hermes on a VPS and have access to a Muse environment that supports hooks.

How it fits with Hermes Agent

Built for Hermes Agent: it appears as a model option that Hermes selects through 9Router.

Requirements: A VPS with 9Router and Hermes running, root or admin access, and a Muse environment that supports hooks

Note: Answers take roughly 20 to 30 seconds per round, and the README also promotes a Muse referral code.

FAQ