Skip to content

Apex

Self-hosted AI agent platform. Multiple models. Persistent memory. Your machine, your data.

Apex on desktop and mobile

Get Started View on GitHub


Why Apex?

Every hosted AI chat sends your data to someone else's servers. If you're working with proprietary code, sensitive documents, client data, or anything you wouldn't paste into a public website — that's a problem.

Apex keeps everything local. Your prompts, AI responses, files, and conversation history — all on your machine. The only external calls are to the AI providers you choose, using your own accounts.


Multi-model, one interface

Use Claude, Codex, Grok, and local models through one unified chat. No new API billing — connect the subscriptions you already have.

Model providers
Provider How it connects What you need
Claude Agent SDK (your Claude Code / Pro / Max subscription) Existing subscription
Codex Codex CLI (your ChatGPT Plus / Pro subscription) Existing subscription
Grok xAI API (pay-per-use) API key from console.x.ai
Local Ollama / MLX Nothing — free, no account

Memory that persists

The AI remembers your projects, preferences, and past conversations across sessions — even across restarts. No manual context management. It just knows.

How memory works

AI personas and group channels

Build specialized agents across different models — Architect on Claude Opus, Developer on Sonnet, Codex on GPT-5.4, Researcher on Grok — each with their own role and memory. Put them in a group channel and they collaborate.

Multi-agent group chat
War Room — multiple agents collaborating in a group channel

See it in action

Five agents from three providers autonomously completed a full security audit — 15 findings resolved in 22 turns for ~$6. One human message started it.


Desktop and mobile

Access from any browser on your network. Native iOS app with push notifications, thinking pills, and streaming indicators.

iOS app with concurrent agents
iOS — Operations responding while Architect, Codex, and Designer work simultaneously

What makes it different

  • Memory system


    Whisper search finds relevant context from past conversations and injects it silently — the AI responds with full history.

  • Your machine, your data


    Everything runs on your hardware. mTLS encryption. No data leaves your network except to providers you choose.

  • Extensible skills


    Slash commands for search, research, delegation, and automation. Build your own or use the built-in set.

  • Admin dashboard


    Manage personas, API keys, model routing, notifications, and security — all from a built-in web dashboard.


Ready?

Follow the Getting Started guide — you'll be up and running in about 10 minutes.

Get Started