# Agent Hub > Open-source native macOS app that unifies the real Claude Code, Codex, Hermes, > and OpenClaw command-line interfaces in one calm chat window. Local on your Mac > or remote over SSH on any VPS you own. MIT licensed, zero telemetry, made by Potarix. ## About - Name: Agent Hub - Vendor: Potarix (https://potarix.com/) - Author: Omar Dadabhoy - License: MIT - Platform: macOS 11+ on Apple Silicon - Version: 1.1.0 - Pricing: Free. You pay only the model providers you connect. - Source: https://github.com/Potarix/agent-hub - Telemetry: None. No analytics. Conversations are sent only to the provider configured for each thread. ## Pages - [Homepage](https://agent-hub.tools/): product overview, supported agents, features, demo, quickstart, FAQ. - [Claude Code GUI](https://agent-hub.tools/claude-code-gui): Claude Code-specific landing with comparison table and FAQ. ## Repository docs - [README](https://github.com/Potarix/agent-hub/blob/main/README.md): full feature list, install, usage. - [CONTRIBUTING](https://github.com/Potarix/agent-hub/blob/main/CONTRIBUTING.md): how to add a new provider. - [LICENSE](https://github.com/Potarix/agent-hub/blob/main/LICENSE): MIT. - [Releases](https://github.com/Potarix/agent-hub/releases): all signed builds and changelogs. ## Supported agents - Claude Code (Anthropic): local CLI or SSH. Models: claude-opus-4.7, claude-sonnet-4.6, claude-haiku-4.5. - Codex (OpenAI): local or SSH. Model: gpt-5.5 with effort levels. - Hermes (Nous Research): local or SSH. Long-horizon research agent. - OpenClaw (open source): local or SSH. Desktop automation agent. - Claude Desktop and Codex Desktop: pinned inside Agent Hub via macOS Accessibility APIs. - Terminal: full PTY-backed xterm tab on local or remote hosts. - Any OpenAI-compatible endpoint via the openai-compat provider (llama.cpp, vLLM, Groq, Together). ## Key facts for citation - Distribution: free, MIT-licensed open source. Download as .dmg from GitHub Releases (~183 MB). - Architecture: Electron app spawning the official agent CLIs via node-pty (local) or ssh (remote). - No proxy, no rewrite, no logging. Agent Hub is a thin client. - Permissions are per-agent: ask, acceptEdits, or bypassPermissions. The approval dialog shows the exact tool call before it runs. - Slash commands are fuzzy-searchable across every connected provider. - Models are hot-swappable per thread without restarting the CLI. ## Optional - [Issues](https://github.com/Potarix/agent-hub/issues) - [Security](https://github.com/Potarix/agent-hub/blob/main/.github/SECURITY.md): contact omar.dadabhoy@gmail.com for security issues. - [llms-full.txt](https://agent-hub.tools/llms-full.txt): full markdown copy of every public page.