Open SEO + GEO toolkit for AI agents

Turn AI agents
into traffic agents.

The open traffic toolkit for agents.

Connect your site, Search Console, and GA4 once. Give Codex, Claude Code, and other agents the context to find opportunities, plan content, ship reviewable changes, and measure impact.

Research · Plan · Write · Optimize · Verify

Install in one command
npx -y aitraffic@latest
View on GitHub
Available for these agents

Built for builders and their agents

Less dashboard. More finished work.

AItraffic turns scattered search signals into one approved, reviewable change.

TerminalOnboarding · v0.8.1
npx -y aitraffic@latest
┌ AItraffic onboarding◇ Environment inspected
  1. 01
    Where should AItraffic be available?│ Codex
  2. 02
    Site URL (optional)│ https://aitraffic.dev
  3. 03
    Google Analytics and Search Console│ Connect Google · opens browser consent
  4. 04
    Choose a GA4 property│ Select from your account
  5. 05
    Choose a Search Console site│ Select a verified site
  6. 06
    Apply this setup?│ Yes
◇ AItraffic added to Codex└ AItraffic is ready

Exact prompts from the CLI. Available properties depend on the signed-in Google account.

Setup stepsLocal first
  1. Choose your agent

    Codex, Claude Code, Hermes or OpenClaw

  2. Add your site

    Start with a crawl and content inventory

  3. Sign in with Google

    Secure browser consent

Safe by design

Google consent stays in your browser.

Credentials stay in your OS vault, never in agents.

WebsiteCrawl starts after setup
Google Search ConsoleChosen during setup
Google Analytics 4Chosen during setup

No automatic publishing. You approve every change.

One workflow, two ways to run it

Open locally. Continue in Cloud.

Start with the open-source CLI. Add scheduling, collaboration, and CMS delivery only when you need them.

Open source

Run it where your code lives.

Your machine, your credentials, your reviewable changes.

Cloud beta

Keep the work moving.

Schedules, team history, approvals, and verification.

Try Cloud beta

Three actions, not 300 warnings.

Every run ends with a short queue you can understand.

Evidence before recommendations.

Observed data stays separate from informed inference.

You approve every change.

Agents prepare the work; you keep control of what ships.