Skip to content
AI agents

Does your agent run SEO and GEO? Plug it into Atyla.

Atyla’s MCP server and REST API give an agent everything it needs: how visible the brand is in AI answers, the exact list of site fixes, the action plan, then a check of the live page. It reads, fixes and verifies. You stay in control.

ChatGPTPerplexityGeminiClaudeMistralGrok

Who it’s for

Autonomous agents

An agent that creates and runs a company, a project or a site can make its SEO and GEO automatic: it reads the measurements, fixes, verifies, with no screen to open.

Teams that hand SEO to a coding agent

Claude Code, Codex or Cursor apply the fixes in your repository or CMS. An action can wait for your approval before anything goes live.

In-house agents

Any client that speaks MCP, or any script that can call a REST API with a key, can read the measurements and the action plan.

What you can ask your agent

Example requests. The agent calls Atyla’s tools, reads the real AI answers, then acts on your site.

  • On which questions are competitors cited instead of us this week?
  • Take this week’s mission, fix the pages and verify them live.
  • Which pages do AI engines read to answer our questions, and which ones cite us?
  • Can AI crawlers read our site?

What an agent can do with Atyla

Read visibility

The share of answers that cite the brand, with its margin of error and trend, engine by engine. The tracked questions, the real AI answers and the competitors cited instead of you.

Know what to do

The action queue in the same order as in the app, this week’s mission, lost questions, and the pages to create or complete.

Fix with the exact list

For each page: every fix the audit asks for, with its current value and its target. Plus the brand’s fact sheet, so nothing gets invented.

Verify the live page

The agent re-reads the published page and sees, fix by fix, what is done. Atyla then measures the effect on AI citations.

How Atyla measures, and its limits

The MCP server’s tools

The agent discovers them on its own. None of them triggers a measurement or a paid job: it can call them freely.

Read the measurements

atyla_overview
Share of answers citing the brand, margin of error, trend, result per AI engine, competitors cited more.
atyla_list_questions
The tracked questions, the share of answers citing the brand and the competitor named most often.
atyla_get_answers
Recent real AI answers to a tracked question, and whether the brand is cited.
atyla_list_sources
The websites AI engines read to answer: media, reviews, comparisons, competitors.
atyla_compare_competitor
The head-to-head with a competitor: where it is cited more, and which of its pages AI reads most.

Pick the work

atyla_list_opportunities
The action queue, in the app’s order, with the reason for each rank.
atyla_weekly_mission
This week’s work: pages to finish, pages to fix, technical issues, questions to cover.
atyla_list_content_gaps
Lost questions, pages to create and pages to expand.
atyla_get_page_signals
Google queries, visits from AI assistants and bot reads, for one page.

Audit the site

atyla_audit_summary
The weekly audit in plain language: be found, be understood, be cited.
atyla_get_page_audit
The exact fixes for a page: criterion, current value, target, why.
atyla_list_technical_issues
Site-wide technical fixes, grouped.
atyla_list_site_changes
What was fixed since the last audit, and what remains.
atyla_get_site_access
The robots.txt rules for each AI crawler, llms.txt and the sitemap.

Act and prove

atyla_get_brand_facts
The facts the brand declared (prices, delays, guarantees), to be used as written.
atyla_list_actions
The action plan: status, page, due date, measured effect.
atyla_start_action
Puts an opportunity in the action plan, before editing the site.
atyla_update_action
Changes the status (needs review, done…) and adds a note.
atyla_verify_page
Re-reads the live page and says, fix by fix, what is done or not.

Plug your agent in, in four steps

The MCP server lives at https://api.atyla.io/api/mcp. It answers over HTTP and expects an API key in the Authorization header.

  1. 1

    Get the Growth plan

    MCP and the REST API are included in Growth and Custom. Keys work from the first payment.

  2. 2

    Create an API key

    In Atyla: Settings › Integrations › Your AI agents. One key per brand, read or write, revocable at any time. It starts with atk_.

  3. 3

    Paste the configuration

    Claude Code, Codex or Cursor: here is the expected format. Replace atk_YOUR_KEY with your key.

  4. 4

    Give it a mission

    The agent lists the tools itself. Start with “review our visibility in AI answers”. Two ready-made commands exist too: fix-site and audit-review.

claude mcp add --transport http atyla https://api.atyla.io/api/mcp --header "Authorization: Bearer atk_YOUR_KEY"

Keep the key secret. If it leaks, revoke it in Atyla and create another.

The REST API, for agents that don’t use MCP

Same key, same Authorization: Bearer header. Base address: https://api.atyla.io/api. The read routes below, plus creating and updating an action. Limit: 120 requests per minute per key.

  • GET /v2/visibility
  • GET /v2/queue
  • GET /v2/actions
  • POST /v2/actions
  • PATCH /v2/actions/{id}
  • GET /v2/lost
  • GET /v2/plan
  • GET /v2/answers
  • GET /v2/sources
  • GET /v2/face
  • GET /v2/audit/plan
  • GET /v2/audit/overview
  • GET /website-audit
  • GET /v2/traffic
  • GET /v2/search-crossing
  • GET /v2/site-access
  • GET /v2/perception/facts
  • GET /v2/workspace
  • GET /v2/brands

The loop, with nobody at a screen

1Measure

The agent sees what AI says

The share of answers citing the brand, with its margin, and the real answers behind every figure.

atyla_overview, atyla_get_answers
2Act

It fixes the site

It takes this week’s work, puts it in the action plan and applies each fix in the repository or CMS.

atyla_weekly_mission, atyla_start_action, atyla_get_page_audit
3Prove

It verifies, Atyla measures the effect

The live page is re-read, the action is marked done, then Atyla compares before and after on the questions it targets. First verdict after 7 days.

atyla_verify_page, atyla_update_action
The right plan

Growth: the MCP server and REST API are included

€199/month excl. VAT, or €1,990 a year. 7-day trial, nothing charged before day 8. API keys work from the first payment, not during the trial. Starter (€89/month excl. VAT) includes neither MCP nor the API.

  • MCP server (Claude Code, Codex, Cursor) and REST API
  • 60 questions tracked every day
  • 6 AI engines: ChatGPT, Perplexity, Gemini, Claude, Mistral and Grok
  • 10 competitors tracked
  • Weekly audit of 1,000 pages
  • Custom from €499/month excl. VAT: several brands

Frequently asked questions

Can an AI agent run a company’s SEO and GEO with Atyla?

It can do everything that happens on the site: read the measurements, pick what to fix, apply the fixes page by page and verify them live. Atyla does not guarantee citations: AI engines choose their own sources. And part of the work happens off the site (media, reviews, forums): the agent sees it in the sources, but Atyla does not do it for you.

Which agents can connect?

Atyla provides configurations for Claude Code, Codex and Cursor. Any other client that speaks MCP over HTTP with an Authorization header works too. Otherwise, the REST API works with the same key.

How does the agent authenticate?

With an API key that starts with atk_, sent in the Authorization: Bearer header. One key per brand, read or write, revocable at any time. Without a key, the server answers 401 and says where to create one.

Can the agent change my site on its own?

Atyla never publishes anything. The agent edits your site, in your repository or CMS, with the rights you give it. It can also set an action to “needs review” so a human checks before it goes live.

What can it write in Atyla?

Create an action in the action plan and update it (status, note). Everything else is read-only. A read-only key cannot write anything.

Could my agent invent figures on my site?

The server’s instructions tell it never to invent a figure, price, delay or testimonial, and to leave a bracketed placeholder instead. It has the fact sheet you declared in Atyla to write with true facts.

Which plan do I need, and what does it cost?

Growth, at €199/month excl. VAT or €1,990/year excl. VAT. Custom, from €499/month excl. VAT, includes it too. Starter, at €89/month excl. VAT, includes neither MCP nor the API. Keys work from the first payment, not during the 7-day trial.

Is there a call limit?

Yes: 120 requests per minute per key. MCP tools read data that is already measured: calling a tool starts no measurement and no paid job.

What does ChatGPT tell your customers this morning?

See it on your own questions, against your competitors, as soon as you sign up.

See my AI visibility7-day trial · From €89 / month excl. VAT after that · Cancel in one click