Working with AI
What the Praxis CLI and skills let your AI assistant do in Facets, how to set them up, and what they add to the assistant.
Praxis is the CLI and skills that let your AI assistant work with Facets. Ask your AI assistant for an outcome in plain language, and it does the Facets work with your own access.
What Praxis is
Praxis has two parts. The Praxis CLI connects your AI assistant to your Facets control plane, and its skills teach the assistant how Facets works. The assistant then works in the same projects you do, and each change it makes ships as a normal Facets release, like one you start yourself.
What you can ask it to do
Most of what you'd do in Facets yourself, from building a module to running a release. What teams ask it first:
Answer questions
"How does checkout reach its database?" It answers from your projects, releases, blueprints and connected repositories.
Design and ship a change
"Add a Redis cache to checkout and release it to dev." It edits the blueprint and runs the release.
Import what's running
"Bring our production databases under Facets." It adopts them without changing them.
Find why something failed
"Why did the last staging release fail?" It reads the logs and configuration, then gives you the cause and the fix.
Praxis adds no approval prompt of its own: your AI assistant's approval settings and your Facets role decide (Approvals and guardrails).
Using Praxis from your own AI assistants
Install the Praxis CLI and log in once. Claude Code, Codex, Cursor, Gemini CLI and Google Antigravity can then do Facets work with your own access.
brew install --cask Facets-cloud/tap/praxis
praxis login --url <your-control-plane-url>On Linux, download the binary instead (Install Praxis CLI). Cloud and Kubernetes commands run on Facets' side, and their credentials stay there.
Cursor reads the skills the Praxis CLI installs for Claude Code, Codex or Gemini CLI on the same machine. With Cursor alone, the Praxis CLI doesn't install skills yet.
In CI, the Praxis CLI signs in with a Praxis API key, which carries its creator's permissions.
What Praxis adds to your AI assistant
Your AI assistant and its model do the work. Praxis adds what the assistant doesn't have on its own:
- Knowledge of Facets: skills for blueprints, modules, releases, imports, and cloud and Kubernetes work.
- Knowledge of your setup: your projects, releases and environments, the repositories you connect, and Infrastructure Graph, which ties a project's code to its Facets resources.
- Limits: it changes only what your account may, and its changes go out as releases, under your approvals and guardrail policies.