Skip to content
Cratis Cratis Cratis

Start using AI with Cratis

A general coding assistant can read C#, but it does not automatically understand Cratis commands, facts, event sources, projections, or vertical slices. Cratis closes that gap in two places: it gives the assistant documented operating tools today, and it is preparing portable coding skills for supported agent ecosystems.

Set up the operating tools

Section titled “Set up the operating tools”

  1. Install the Cratis CLI

    Follow the CLI getting-started guide for your platform, then open a terminal in your Cratis project.

  2. Generate the agent-facing CLI context

    Terminal window
    cratis init

    The command writes the CLI context and supported host instructions for the project. Your assistant can now discover documented Chronicle inspection and diagnosis commands instead of inventing shell syntax.

  3. Refresh after a CLI upgrade

    Terminal window
    cratis init --refresh

    Refreshing keeps the assistant’s command catalog aligned with the CLI version you actually run.

  4. Connect Chronicle MCP when your host supports MCP

    Use the Chronicle MCP setup guide to connect an approved host to a running Chronicle store. Review every requested capability before enabling mutating operational tools.

You now have the operating side of Cratis AI: the assistant can use documented CLI or MCP surfaces to inspect Chronicle with the same vocabulary you use in the application.

Use coding skills in a Cratis-owned repository

Section titled “Use coding skills in a Cratis-owned repository”

Cratis maintainers can use the repository-local AI setup already present in Cratis repositories. Start your agent from the repository root so it reads AGENTS.md and the repository’s current host adapter. Ask for the outcome in plain language—for example, “add a projection,” “model this workflow,” or “review these changes”—and let the matching skill apply the repository profile and quality gates.

See Using AI as a Cratis maintainer for the complete workflow and the rules for project context, verification, and shipping.

Set up coding skills in your own repository

Section titled “Set up coding skills in your own repository”

Portable public skills are not released yet. Cratis is validating one canonical skill tree across Agent Skills, Claude Code, Codex, GitHub Copilot, Cursor, Kiro, Junie, Gemini CLI, Pi, and npm packaging before opening installation.

Until an immutable release is published:

  • do not copy .ai, .claude, .github, or skill folders from another repository;
  • do not install the fixture-only distribution repository;
  • keep your project-specific facts in your repository rather than in shared skills;
  • use cratis init and Chronicle MCP for the currently documented operating workflow.

The ecosystem support matrix shows the generated format and current verification level for every host. The trust and distribution model explains why Cratis uses versioned generated artifacts instead of copying one repository into another.

Choose your next step

Section titled “Choose your next step”
Cratis maintainers

Work effectively with the repository-local rules and skills already used by the Cratis team.

 Agent ecosystems

See Claude, Codex, Copilot, Cursor, Kiro, Junie, Gemini, Pi, and npm status in one place.

 Trust and distribution

Understand generated packages, approvals, canaries, rollback, and why propagation stays frozen.