Automatic optimization

Your AI marketing director, on autopilot

Rempa measures what ChatGPT, Gemini, Perplexity, Copilot, Claude and Google's AI say about your brand, and decides what to fix. Connect your AI agent and it applies the plan in your own tools, records what it did, and Rempa checks whether the answers changed. You approve what goes public.

Recommendations are included in every paid plan, from $75 a month.

What the autopilot does

  • Finds what to fix: AI crawlers blocked, pages missing, websites where your competitors are cited and you are not.
  • Ranks it by what is at stake: the monthly searches behind the questions you lose.
  • Hands the plan to your agent through the MCP connector, with the evidence for each item.
  • Records each step: in progress, done or dismissed, straight from the agent.
  • Checks the result at every update, and closes what no longer shows.
The loop

Measure, decide, act, verify

Most GEO work stalls between the report and the fix. The autopilot closes that gap: the same plan goes from Rempa to your agent and back, update after update.

01 Rempa

Measure

Your buyers' questions go to the AI models you choose, from clean sessions, in your market. Rempa records whether you are named, in which position, in what tone, and which pages each answer cites.

02 Rempa

Decide

After each update, Rempa rebuilds the recommendations: technical fixes, pages to publish and websites to get listed on, ranked by the searches at stake, each with its evidence and steps.

03 Your AI agent

Act

Claude Code, Cursor or any agent that speaks MCP reads the plan, fixes your website, drafts the missing pages and prepares the outreach, with the access you give it.

04 Rempa

Verify

The agent marks each recommendation in progress or done. The next updates show whether the answers changed; a problem that no longer shows drops out of the plan.

What gets applied

Automatic where it is safe, reviewed where it matters

RecommendationWhat your agent doesYour part
AI crawlers blocked in robots.txtAllows the search crawlers of ChatGPT, Perplexity and the others, and keeps the rules you wantMerge the change
No llms.txtWrites one from your site: who you are, your products, your key pagesMerge the change
No Organization structured dataAdds it to your home page with your name, logo, website and profilesMerge the change
A page to publishDrafts it from the questions you lose, the brands named instead of you and the sources citedReview, then publish
A website to get listed onLists the pages to target and prepares the listing details and messagesSend them
Your pages rarely citedReviews the pages that answer your buyers' questions and proposes edits that put the answer firstApprove the edits
Set up

Running in three steps

  1. Track your brandCreate a project from your website: Rempa writes the prompts, finds your competitors and runs the first update in minutes.
  2. Connect your agentIn Claude Code or Cursor, add the Rempa MCP server with an API key. In claude.ai or ChatGPT, add it as a connector and allow updates.
  3. Give it the jobAsk it to apply the recommendations after each Rempa update, weekly, or as part of your team's routine.
Claude Code claude mcp add --transport http rempa https://api.rempa.ai/mcp \ --header "Authorization: Bearer rmp_your_key" You › Apply this week's Rempa recommendations for acme-us. Agent › 4 recommendations, by searches at stake ✓ robots.txt lets OAI-SearchBot in → change ready to merge ✓ Organization schema on the home → change ready to merge ✎ Page "Acme vs Orbita" drafted → waiting for your review ✉ G2 listing: review requests ready → yours to send Rempa updated: 2 done, 2 in progress.

You stay in control

  • Your websiteRempa never touches itRempa measures, recommends and records statuses. Your agent works in your tools, with the access you grant it.
  • PermissionsRead, or read and updateA connection allowed for reading only can't change anything. Updating a recommendation's status needs your explicit consent.
  • TeamViewers can't connect agentsOnly the account's owner and members can create API keys or connect an app; you can disconnect any of them in Settings → API.
FAQ

Autopilot questions

Does Rempa change my website?

No. Rempa measures the AI answers, builds the recommendations and records their status. The changes are made by your own AI agent, in your own tools, with the access you give it: a code change you merge, a draft you publish, a message you send.

Which AI agents work with the autopilot?

Any agent that supports the Model Context Protocol (MCP): Claude Code, Cursor and VS Code with an API key, and claude.ai or ChatGPT as a custom connector with your Rempa sign-in. To apply website fixes, the agent also needs access to your website's code or CMS.

Which plan do I need?

Recommendations are part of every paid plan: Starter ($75 a month, updated every 3 days), Pro ($195, every day) and Enterprise ($395, every day). The API and the MCP connector are included in all of them.

Is it safe to let an AI agent publish pages?

We recommend that it doesn't publish on its own. Technical fixes are small and checkable, so they go out as changes you merge. Pages and outreach carry facts, prices and your brand's voice: the agent drafts them, a person reviews them.

How do I know it worked?

Each recommendation keeps its status and date. The following updates show your visibility, position and citations on the questions it was about; when a problem no longer shows, the recommendation drops out of the active plan.

Can agencies run it for their clients?

Yes. Each client is a project, and the agent works project by project: it reads each client's recommendations and records progress in the right one, which also gives the agency a log of the work done.

Put your AI visibility on autopilot

Start with the free trial, then connect your agent when your first recommendations are ready.