Semrush MCP: AI Visibility and Keyword Gaps
Quick answer: Connect Semrush to a coding agent over MCP with an API key, read AI visibility and citation numbers from the dashboard, then ask the agent for an implementation plan on keyword and backlink gaps. Review that plan before you ship anything. This page follows the 6 August 2026 recording. It is not a current Semrush price list.
This recording is not a paid promotion. Mike said the YouTube description link for Semrush was an affiliate link. This page uses the public product site https://www.semrush.com with no tracking parameters. For a sponsored Rank Prompt scan plus Claude Code MCP, see Rank Prompt AI Visibility: Scan, MCP and Fixes. For a free VisiScan domain audit, see VisiScan AI Visibility: Free Scan and Report.
What this recording shows
Mike walks through Semrush as an SEO and AI visibility tool, then pulls the same data into coding agents over MCP. The domain he types at the start is Tube Analytics. He says you create a free account first, and that he would upgrade to a paid plan to get MCP access for those agents. If MCP itself is new, start with What is MCP? The Model Context Protocol explained.
On camera the agents are OpenCode and Hermes Agent. The job in this session is to close keyword gaps and backlink gaps versus competitors, using live Semrush data rather than guesses. Another MCP SEO path, with per-request costs you can check, is DataForSEO MCP: build an AI SEO agent in Claude Code.
1. Read the AI visibility dashboard
The Semrush dashboard in this take shows overall AI visibility at 14 out of 100 for his site. He says he would have expected that number to be higher. Total citations across the listed models are about 428, covering ChatGPT, Claude, Gemini, Perplexity and others named on screen.
He opens the listed items, the ChatGPT output for one of them, and the sources that answer recommends. Those figures are from 6 August 2026. They are not a later rerun. For crawler hits versus real visits on a different tool, see Trakkr and Codex: AI Visibility, Crawlers and Visits.
2. Connect the Semrush MCP
He says the dashboard is useful, and that MCP is more useful because a coding harness can call the same data. To start, search for Semrush MCP and read the documentation on that page. The recording names one required piece: an API key. Generate that key, then tell the coding harness to add the MCP, as he already has in this session.
The video does not speak an install command or a config file, so this page does not invent one. Once connected, he says the agent can read AI visibility, keyword gaps, backlink gaps and competitor data from Semrush. Confirm the server in your own client before you approve writes.
3. Turn keyword gaps into a plan
He has already asked the agent to create an implementation plan from the Semrush data. The plan states what it found. He then says:
Implement this plan.He expects that run to close keyword gaps. In this session OpenCode is on the DeepSeek V4 Flash free model, which he describes as 100% free, so he is not spending model credits on that step. Model access is separate from the Semrush plan. Review the proposed page changes before they go live. A report-then-review loop for on-page SEO is in Internal Link Audit with Codex: A Reviewed SEO Workflow.
4. Pull backlink gaps and run a script
On 6 August 2026 he said he was on a $300 per month Semrush plan. The recording also showed a 7-day free trial that included access to the data on screen. Check Semrush for current prices. Do not treat those figures as a live quote.
In Hermes Agent the work is backlink gaps, places he is missing versus competitors. The agent writes a script he can run to pull those gaps from Semrush. During the run it tries to fix a backlinks matrix so API exhaustion errors throw cleanly, because the session is already hitting credit throttling. He still says the useful part is live data rather than guesses. Ahrefs is named as another tool that, in his view that day, tended to cost more than Semrush.
5. Mentions, outreach and disavow
From Semrush he also reads organic traffic, referring domains and backlinks. He says the site then had about 19,000 backlinks. Domain overview in this take also shows citation counts, cited pages and mentions per LLM: ChatGPT 29.4% of mentions, Gemini about 35%, AI mode 23%. Topic opportunities on screen were about 863,000 prompts he might want to rank for. One example query he opens is YouTube and video streaming platforms, with a ChatGPT-style response next to it.
Outreach in this recording is a separate agent path. He uses Agent Mail to send backlink requests to high domain-authority sites, and says Semrush MCP can supply data for that work. Some existing backlinks look low quality on camera, so he talks about disavowing them in Hermes Agent and uploading the list to Google Search Console. That is what he did in the video, not a measured domain-authority result.
For a self-hosted SEO stack with its own data bill, see OpenSEO review: setup, data costs and practical limits. For GrokBot audit bots that still need a review step, see GrokBot for SEO: Audit Bots, Data and Reviewed Fixes.
Watch the walkthrough
FAQ
Was this Semrush video sponsored?
No. The recording was not marked as a paid promotion. Mike said the YouTube description link for Semrush was an affiliate link. This page uses the public product site with no tracking parameters.
What Semrush plan and trial did the recording show?
On 6 August 2026 the recording showed a $300 per month plan and a 7-day free trial that included access to the data on screen. Those were what that date showed, not a current price list.
Which coding agents used the Semrush MCP?
OpenCode and Hermes Agent. Mike said he was pulling Semrush data in those harnesses to close keyword and backlink gaps.
Did the agent finish the keyword-gap plan without errors?
No. After he said implement this plan, the recording still hit API exhaustion and throttling while a Hermes Agent script pulled backlink gaps. Review the plan and the API usage before you treat the run as done.
Sources
- Video: Stop Wasting Money on SEO – Do This Instead (6 August 2026)
- Semrush: https://www.semrush.com
- Model Context Protocol: https://modelcontextprotocol.io/
You may also like
Build the next one with AI Income Lab
Step-by-step courses, reusable templates, and a community that answers when you get stuck.
Members get the Claude Code & Skills and Codex Tutorials & Courses lessons.
- Courses for builds like this one
- Copy-ready templates
- Ask questions while you build
- Weekly coaching on VIP
