vertex-proxy: model tool for Hermes Agent

Local Anthropic, Gemini and OpenAI-compatible proxy that routes requests to Google Cloud Vertex AI

What vertex-proxy does

vertex-proxy is a small local proxy, written in Python, that lets tools speaking the Anthropic Messages API, the Gemini API or OpenAI Chat Completions talk to Google Cloud Vertex AI. It listens on 127.0.0.1:8787, refreshes the short-lived OAuth access tokens that come from a service-account key, and translates request shapes to Vertex publisher-model endpoints, so clients that expect a static bearer header need no changes.

Routes include /anthropic/v1/messages, /gemini/v1beta/models/{m}:{action} and /openai/v1/chat/completions, plus /v1/models listings and a /health check. The OpenAI route reaches Claude, Gemini and partner models such as Kimi, GLM, MiniMax, Qwen and Grok on Vertex. Streaming works on all routes. The README names Hermes Agent, Claude Code, opencode, Cline and Continue.dev as example clients.

Key features

Anthropic Messages, Gemini generateContent and OpenAI Chat Completions endpoints
Automatic service-account token refresh for Vertex AI
Streaming on all routes with no read timeout on the upstream client
OpenAI-compatible route covering Claude, Gemini and partner models on Vertex
Model listing endpoints and a /health check
Local-only binding to 127.0.0.1:8787

When to use it

Using Hermes Agent with Claude or Gemini models billed through a Google Cloud project
Spending GCP credits or using Vertex quotas from tools that expect public API URLs
Giving several clients one local base URL for Vertex AI

Who it is for: Developers with a Google Cloud project who want existing agent clients to use Vertex AI models without code changes.

How it fits with Hermes Agent

It is a general proxy, and Hermes Agent is one of the clients the README lists among Claude Code, opencode, Cline and Continue.dev.

How to install vertex-proxy

These commands are copied from the project's README. Check the repository for the latest steps before you run them.

pipx install vertex-proxy
uv tool install vertex-proxy
uvx vertex-proxy

Requirements: Python 3.11+, a GCP project with the Vertex AI API enabled, and a service-account JSON key with roles/aiplatform.user

FAQ

What is vertex-proxy?

vertex-proxy is a local proxy that translates Anthropic, Gemini and OpenAI-style API calls into Google Cloud Vertex AI requests. It handles service-account authentication so clients do not have to.

Does vertex-proxy work with Hermes Agent?

Yes, as a client. The README lists Hermes Agent among the tools you can point at the proxy's local base URL to use Vertex AI models.

What do I need to run vertex-proxy?

You need Python 3.11 or newer, a GCP project with the Vertex AI API enabled, and a service-account JSON key with the roles/aiplatform.user role. Set the key path and project ID, then start vertex-proxy.