Governed AI coding, ready the moment it installs.
One self-contained install puts cortex code in your shell with OMP pre-configured, the Sema runtime + compiler, and the full Cortex harness — skills, rules, agents, and templates. Works locally with your own model key; connect a Cortex Cloud account only if you want one.
$ curl -fsSL https://cortex.a2olabs.com/cloud/downloads/install.sh | sh ✓ cortex, sema, omp installed → ~/.cortex/bin $ cortex code ▸ governed session · policy=developer · omp
Already installed? One command.
cortex update fetches the latest authenticated release for this OS and architecture, downloads only changed layers, validates the signed staged runtime, and swaps it atomically. Profiles, tokens, model providers, and local configuration remain in place; cortex update --rollback restores the previous bundle.
cortex updateAutomation: cortex update --check. Signed service contract: /update/manifest.json.
Pick your platform
The one-liner detects your OS and architecture, installs everything to ~/.cortex, and wires your PATH — no admin, no dependencies to chase. Only platforms present in the published release manifest are enabled.
No account required — Cortex runs locally with your own model key; sign in only to connect a Cortex Cloud tenant. Verify downloads against manifest.json (SHA-256). Version 0.1.0.
What's in the box
Everything you need for a governed agentic workflow — nothing you have to configure.
cortex code
The governed harness pass-through — policies, approvals, audit, and a smart status bar around your coding agent.
OMP, pre-configured
The OMP coding agent, wired to route every tool call through Cortex governance out of the box.
Sema language
The Sema runtime + compiler on your PATH, with the sema-language skill and check hook — author neurosymbolic code from day one.
The full harness
Skills, rules, agents, instructions, and project templates — the same battle-tested harness Cortex uses.
Zero config
Bundled runtimes mean no Node/Python to install and no distro package manager touched. It just runs.
Your models
Bring an OpenAI/Anthropic/GLM key, point at any OpenAI-compatible endpoint, or use a Cortex-provided model via your account.
Need multi-tenant governance, SSO, observability, shared memory, and one-click deploys?
Explore Cortex Cloud →