Welcome to Cial

Cial is the platform for self-editing AI workspaces. It runs coding agents — harnesses like Claude Code or Kimi — and hands them the context, tools, and a built-in deploy pipeline. The key part: those tools include the ability to reshape Cial itself. The agent can rewrite the platform's own code and redeploy it, live, while you watch.

So Cial isn't a fixed product — it's the substrate the agents run on. The harnesses, tools, and features in these docs are the current setup, not a ceiling. When something's missing, you don't file a request and wait — you have the agent build it into the platform. Cial manages the models and harnesses and gives any agent the keys to self-modify. "Cial doesn't do X" is almost always just "Cial doesn't do X yet."

The one idea

The harnesses, tools, surfaces, and integrations you'll see in these docs are the current implementation — not the point. What's permanent is the self-edit loop: the ability to change any of it. The tech stack is disposable; the capacity to reshape it is the product.

This is why Cial is the ultimate integrator — because any agent running on it can reshape the platform, it can connect to almost anything: a new model provider, your internal APIs, a bespoke workflow, a tool nobody has built yet. The limit isn't a feature list; it's what you can describe.

What you can reshape

Start here