Hermes Cognitive Bridge

Hermes Cognitive Bridge creates a local integration layer: a Chromium extension extracts context from ChatGPT Web, a Python gateway authenticates and routes requests to Hermes CLI, and Hermes executes with full approval mechanisms intact. Zero cloud dependencies.

Developers constantly copy/paste between ChatGPT Web and local AI agents. Existing bridges either require cloud services or compromise security. This keeps everything local.

The repository evidence points to these observable boundaries and technologies:

• Manifest V3 Chromium extension (content script + background)
• Python FastAPI gateway with JWT authentication
• Hermes CLI subprocess management
• WebSocket for real-time bidirectional communication
• Secure token storage in extension storage

• ChatGPT Web DOM extraction for conversation context
• Extension-to-gateway authenticated WebSocket channel
• Gateway-to-Hermes CLI subprocess orchestration
• Hermes approval prompts surfaced in extension UI
• Conversation history sync between web and CLI

• ChatGPT Web DOM instability across updates
• Manifest V3 service worker lifecycle limitations
• Secure local authentication without user friction
• Hermes CLI output parsing for structured responses

• Extension loads and injects into ChatGPT Web
• Gateway authenticates and routes to Hermes CLI
• Hermes approval prompts appear in extension
• End-to-end flow: web → extension → gateway → Hermes → response

Local AI bridges need resilient DOM selectors and graceful degradation. Manifest V3 requires rethinking background script patterns.