AI Agents
contextlint offers two integration paths for use from AI agents. From hosts like Claude Code, Cursor Agent, Cline, Codex, Gemini CLI, and GitHub Copilot, you can fold document-integrity checks into your conversation.
Two integration paths
Section titled “Two integration paths”| Path | Protocol | Main hosts | Distribution |
|---|---|---|---|
| MCP server | Model Context Protocol | Claude Desktop / Cursor / Cline / Codex, etc. | @contextlint/mcp-server package |
| Agent Skills | agentskills.io spec | Claude Code / Cursor Agent / Codex / Gemini CLI / GitHub Copilot, etc. | gh skill install from a GitHub repo |
The two don’t compete — pick whichever suits the goal. MCP is a protocol for AI to call contextlint’s features directly, while Skills are workflows that tell AI what you want done in natural language.
Which to choose
Section titled “Which to choose”- “I want AI to run lint” / “I want AI to read the document graph” → MCP server
- “I want AI to set up contextlint” / “I want AI to fix violations” / “I want AI to analyze impact” → Skills
If your host supports both (Claude Code, etc.), you can use them together. A Skill may call MCP tools internally as well.
What’s in this section
Section titled “What’s in this section”- MCP server —
@contextlint/mcp-serversetup and the five tools it exposes - Agent Skills — using
gh skill installand the list of supported hosts - contextlint-init Skill — let AI set up contextlint in a repository
- contextlint-fix Skill — let AI fix detected violations
- contextlint-impact Skill — let AI analyze the impact of a file change
Related
Section titled “Related”- For the shortest path to setup, see Quick Start — AI integration. This section covers the detailed specs and use cases for each Skill / tool.