Secure AI agents that earn your trust
Apex is a security-first AI agent runtime with cryptographic identity, spawn authorisation tokens, adaptive memory, and a self-hardening immune system.
What makes Apex different
Cryptographic Agent Identity
Every agent gets an Ed25519 certificate. Every action is signed and verifiable. You always know which agent did what, and you can prove it.
Spawn Authorisation Tokens
High-risk actions require a signed token from the operator. Not a config flag, a real-time, cryptographically-verified authorisation that expires after use.
Guardian Agent
Ships as standard. Monitors system health every 4 hours. Cannot be deleted. Cannot be silenced without authorised approval. Read-only: it watches, it never touches.
Adaptive Memory
Agents learn across sessions. Corrections become permanent. Casual mentions fade naturally. Activity-relative decay means idle agents don't lose knowledge.
Model Agnostic
Works with Anthropic, OpenAI, or both. Configure a primary and a fallback. Not locked to any vendor. Your agents, your choice of brain.
System Snapshots
Full system backup with one command. Security material verification, audit chain integrity, Guardian workspace checks. Restore with confidence.
Migration Engine
Move from OpenClaw to Apex in minutes. Config remapping, workspace migration, credential extraction. Migrated skills are quarantined automatically until you approve them.
Apex Immune No one else has this
An adaptive integrity defence system. When something goes wrong, Apex captures the evidence, restores from a clean state, diagnoses what happened, and generates defensive rules that prevent the same class of attack from ever succeeding again. The system gets stronger every time it's attacked. Shield rules start under operator control and progressively earn autonomy as they prove reliable.
Setup in 5 minutes
# Download and extract tar -xzf apex-v0.9.tar.gz && cd apex-runtime # Install dependencies pnpm install && pnpm link --global # Run the guided setup wizard apex onboard ╔══════════════════════════════════════╗ ║ Welcome to SolNet Apex ║ ║ Secure AI Agent Runtime ║ ╚══════════════════════════════════════╝ ? What's your name? Your name ? Which LLM provider? Both (Anthropic + OpenAI) ? Name your first agent: Your primary agent name ✔ Configuration saved ✔ Agent workspace created ✔ Guardian activated ✔ Certificates generated ✔ Memory seeded # Start the gateway apex gateway run # Open WebChat in your browser http://localhost:18789?token=your-auth-token
Requirements
Early Access
Apex is in pre-release testing with a small group of experienced AI practitioners.
Looking for people who will push it, break it, and tell us what's missing.
What we want to know
- Does the security model feel robust, or theatrical? Try to break it.
- Where does the setup experience frustrate you? What took too long or felt unclear?
- How does the memory system behave after 10-20 conversations? Does the agent actually learn?
- What's missing that would make you switch