An AI SEO agent that already has your site’s project data
This is not ChatGPT with a pasted CSV. It is an in-app SEO research agent with live project tools, project memory, editable skills, and MCP so your own assistant can use the same suite. It does not invent volume, difficulty, traffic, ranks, or backlinks.
You slipped on 3 keywords in the last three weeks, all on the same page.
Two competitors published stronger pages and picked up links to them. Your page hasn't been updated since spring. Here's where I'd start:
- 1. Refresh the page and answer the two questions both competitors cover.
- 2. Go after the 4 referring domains they both earned.
Example prompts
- Why did we lose rankings on this set of keywords?
- What should we write next for this site?
- Who are our AI competitors for this category?
- We rank but get almost no clicks. What’s broken?
MCP short path
- Connect Claude Code, Codex, or another MCP client
- Point it at the project-scoped MCP URL
- Use the same project tools without copy-pasting data into a disconnected model
What you can do
Can I ask plain-English questions about my SEO data?
You get a chat interface on the project. The agent reads Search Console and rankings when those are connected, plus research, audits, backlinks, domain work, opportunities, Visibility, analytics, and the rest of the project tools as those surfaces exist. Tool use is progressive. It is not every tool loaded on every turn. You ask in plain English; it goes get what it needs.
Does it just dump charts, or tell me what to do?
It is supposed to translate research into a next move. When you ask what to work on, it should lead with Do Next, not a flat pile of charts. It can also hand work to opportunity flows when that is the right path. Plain answers first. Prioritized work second. That is the point of an agentic SEO assistant for agencies, not a generic ChatGPT prompt with no project grounding.
Does the agent remember my project between sessions?
Project memory is durable for that project, and you can edit or delete what it keeps. It also uses project context for business, customer, and market so the agent is not starting from zero every Monday. That is the difference between “new chat” and “same site, same story.”
Can we customize how the agent works for our team?
Skills are pre-seeded and fully editable. You can edit, enable, disable, and reset defaults. Packaged workflows cover things like setup, research, competitive work, link prospecting, GEO site changes, CTR cleanup, orphan rescue, and ship-and-verify style passes. Treat them as starting points your team can tailor, not as a locked black box.
Can my own AI assistant use the same SEO tools?
Yes. We ship an MCP server so Claude Code, Codex, and other MCP clients can call the same project tools after OAuth consent. Each project gets a project-scoped MCP URL so client A does not bleed into client B. There is a tool catalog of what agents can call, and installable external skills for teams that want workflows outside the in-app chat.
Will it invent rankings or traffic numbers?
No. Product rule: never invent volume, difficulty, traffic, rankings, or backlinks. It should check redirect and non-canonical issues before leaning into optimize advice. After recent edits it can say it is too early to judge. Paid originality scans stay human-approved. If you wanted an agent that makes up comforting numbers, this is not that.
What you see
- Plain-language answers grounded in project tools
- Prioritized next steps when “what next” is the question
- Keyword, competitor, and ranking diagnoses
- Markdown you can copy or download for handoff
Want the operator-level detail?
Open the docs for SEO Agent workflow, limits, and behavior notes.
Questions
How do agencies keep client memory separate?
One project per client or property. Memory is per project. Multi-project setup keeps agency work from turning into one shared brain.
Do I have to use MCP, or is the in-app agent enough?
In-app alone is a full path. MCP is for teams that already work in Claude Code, Codex, or another MCP client and want the same tools there.
Does everything the agent does cost credits?
Paid actions show cost before you run them. Many reads of connected first-party data stay free. Qualitative only here. See /pricing for how cost works. No sample numbers on this page.
Can I have an AI SEO agent work a keyword opportunity end to end?
You can hand opportunity work to the agent and keep research on the same project tools. Paid or sensitive steps stay on honest gates.
Works well with
All featuresOpportunities
Three opportunity lanes plus the path for what to work next.
Visibility Engine
See if you show up in AI answers, then prioritize the next fix and remeasure.
Keyword Research
Find, compare, and save keyword opportunities for the project.
Reports
Decision packs, weekly digests, and one-zip project export.
We are still in private testing
If you want in when spots open, request early access. If you have a sharper question first, send a message.