Build with an AI agent
Sign in with praxis to install the Facets skills into your AI coding agent, then build and release infrastructure from your terminal using raptor, with no cloud credentials on your laptop.
Facets is built to be driven by an AI coding agent. Signing in installs a set of Facets skills into whichever agent you already use, so it knows the blueprint model, the module contract, and which commands to run. You describe the outcome, the agent does the work with raptor.
Sign in once
praxis login
raptor loginpraxis login installs the Facets skills into every AI host it detects on your machine. It writes to ~/.claude/skills/ and to ~/.agents/skills/, the shared location that other agents read, so this works with Claude Code, Codex, Gemini CLI and anything else following that convention. It also fetches your organisation's skill catalogue, so the skills you get reflect your control plane rather than a generic set.
raptor login authenticates the CLI against the control plane. Confirm both:
praxis status
raptor whoamiTo see what was installed and where it landed:
praxis list-skillspraxis login installs your organisation's skill catalogue from the control plane. Raptor also embeds its own CLI skills, which come from the binary rather than the server and are installed separately:
raptor install-skills --agent claude # also: gemini, codexMost people want both. See Commands for that side.
MCP tools run server-side under your organisation's managed credentials. Your laptop does not need AWS, Kubernetes or Terraform credentials for the agent to work.
What the agent can do
Once the skills are installed, your agent understands the Facets workflow and reaches for the right command without being told the syntax. In practice that covers:
- Populating a new control plane, which is importing a project type.
- Adding and configuring resources in a blueprint, and setting per-environment overrides.
- Planning and releasing changes, then reading the logs when a release fails.
- Authoring and changing IaC modules, including validating and publishing them.
You still own the decisions. The agent proposes commands and you approve them, and everything it does goes through raptor against the same control plane you would use by hand, so there is nothing it can do that you could not audit or repeat yourself.
Keeping skills current
Your organisation's catalogue changes over time. Refresh without a full re-login:
praxis refresh-skillsNext steps
- Import a project type if you have not yet, since a new control plane has no modules.
- Create a project and add resources to its blueprint.
- Building a module when you need a capability your catalogue does not cover.
Import a project type
A new Facets control plane starts with no modules. Import an official project type to load the modules and output types for your cloud in one command, then adapt them to your organisation.
Setup Platform
Platform setup: projects, environments, secrets & variables, resources, and modules. Everything a platform team configures once so application teams have a clear path to production.