Coding Agents

RootCX uses two complementary open integrations:

• MCP gives the agent authenticated access to project context, manifest validation, and governed data operations.
• The RootCX Agent Skill provides the implementation workflow and detailed RootCX references.

Connect Claude Code

claude mcp add --transport http rootcx https://<tenant>.rootcx.com/mcp

Run /mcp inside Claude Code and complete the RootCX browser sign-in. Tokens are stored and refreshed by Claude Code.

Install the RootCX skill

npx skills add rootcx/rootcx-skills -g -y

This installs one rootcx skill that uses progressive disclosure. It loads only the reference required for the current task: data and manifests, UI, backend workers, AI agents, or integrations.

The skill follows the Agent Skills open specification. Client support and installation UI vary, so only install it in a skills-compatible coding agent. Chat-based MCP apps can use the server's tool descriptions and do not require a separately installed skill.

First-app workflow

1. The agent calls get_project_context.
2. It asks for the outcome, intended users, and information to manage.
3. It restates the user story before writing code.
4. It runs local checks, calls validate_manifest, and resolves every validation error.
5. It requests approval, then runs rootcx deploy locally.
6. It returns the live app URL.

The CLI is the only application build and deployment path. The skill installs and uses it in the background when local command execution is available.