AI Agent

AI Agent

RankMySEO includes an optional AI agent layer for conversational SEO assistance.

Package

@rankmyseo/agent — tool definitions, system prompts, and orchestration logic.

Tools (MCP-style)

The agent can invoke server-backed tools:

Tool	Description
scan_url	On-page analysis of a URL
generate_meta	Meta title/description generation
generate_schema	Schema.org JSON-LD (Article, Product, FAQ, Breadcrumb, Organization)
list_keywords	Tracked keywords for project
get_rank_history	Rank snapshots for a keyword
run_audit	Trigger site audit
list_blog_posts	Blog posts (when enabled)
create_blog_post	Create draft post

Tool schemas live in packages/agent/src/tools/.

React integration

import { useRankMySeoChat } from "@rankmyseo/react";

function ChatPanel() {
  const { messages, send, streaming, error } = useRankMySeoChat();

  return (
    <div>
      {messages.map((m) => (
        <div key={m.id}>{m.role}: {m.content}</div>
      ))}
      <button disabled={streaming} onClick={() => send("Scan my homepage")}>
        Send
      </button>
    </div>
  );
}

Server route

Agent chat is exposed via the server package (streaming SSE or chunked responses depending on adapter). Configure LLM credentials in environment — not bundled in OSS defaults.

Environment

Typical variables (exact names in server config):

OPENAI_API_KEY=sk-...
# or compatible OpenAI API endpoint

Without credentials, agent routes return appropriate errors; core rank/audit/blog features work independently.

Extending tools

1. Add tool definition in packages/agent/src/tools/
2. Register handler in agent router
3. Update system prompt with tool description
4. Add tests with mocked LLM responses

OSS scope

The agent is optional. Rank tracking, audits, scan, meta, and blog do not require AI. This matches the product goal: free-first SEO tooling with AI as an enhancement.

See Development for running the agent locally.