The short answer
Encited has shipped an MCP server, or more precisely six of them, that let an AI agent read your Search Console data, your AI visibility tracking, your site audits and your competitor research, then act on what it finds without you leaving the chat.
The interesting part is not that it exists. It is how much of the workflow is prewritten. Their launch page carries around thirty ready-made prompts, each one paired with a named content agent and a Run in Claude button. You are not being handed an API and wished luck.
| Question | Answer |
|---|---|
| What is it? | Six MCP servers over Encited’s existing platform data |
| Which clients? | Claude, Claude Code, ChatGPT, Cursor, Lovable |
| Separate price? | No. It reads whatever your Encited plan collects |
| Cheapest plan that includes it | Essential, $49/month billed annually |
| Is the flagship workflow on that plan? | No. Fan-out queries start at Business, $149 |
| In Claude’s directory? | Yes, as a Community connector |
We have not run it. Everything below was read off Encited’s own pages and Claude’s connector directory on 10 September 2026. There is no score on this page and no verdict, because we have not used the product, and the section at the end says exactly what that leaves unanswered.
What an SEO MCP actually changes
An MCP server is a standard way of handing an AI assistant a set of tools and a live connection to your data. It matters here for one specific reason, and it is worth being precise about it rather than calling it automation.
The old shape was: open the dashboard, read a chart, form a theory, export a CSV, paste it somewhere, write the fix. The chart and the fix lived in different places and a human carried data between them. That carrying is most of the work, and it is the part that silently does not happen on a busy week.
With an MCP the assistant holds both ends. It queries the data, reasons over it, and writes the change in the same conversation, and in Cursor it can apply that change to your codebase directly. The saving is not the analysis. It is the handoff.
Which is also the risk
An agent that can read your Search Console and rewrite your titles in one motion is an agent that can rewrite your titles based on a misreading. The prewritten prompts help, because a prompt someone tested is better than one you improvised, but the general rule holds: the further an agent gets from a human reading the output, the more the quality of the underlying data matters.
Saru says: The prompts that end in a plan are safer than the ones that end in a change. Start with the ones that hand you a ranked list and let you decide, and only automate a workflow after you have watched it be right a few times.
A worked example, since the abstraction hides the point
Take one of their published prompts: pull the pages sitting in positions four to thirty, match each against the titles and H1s from your last crawl, and rewrite the ones that miss.
Done by hand that is four tools and most of an afternoon. Search Console for the positions, a crawler for the titles, a spreadsheet to join them, then a writing session. The joining step is the tedious one and it is why the job gets postponed. An agent holding both data sources does the join as a side effect of being asked the question, and the afternoon becomes a prompt and a review.
Notice what has not changed, though. Something still has to decide whether the rewritten title is better, and that something should be you. The tool has compressed the retrieval, not the judgement.
The six servers
Encited splits the offering by job rather than by data source, which reads better than the underlying product structure probably is, but it does make the choice easier.
| Server | What it is for | Most useful prompt on the page |
|---|---|---|
| Search Console | Mining up to a year of GSC history | CTR gaps, cannibalisation, decaying pages and snippet targets, returned as a ranked fix plan |
| AI Visibility | Whether AI engines mention you, and the searches behind their answers | Cluster fan-out queries into a page by page content plan with off-site citation targets |
| Technical SEO Audit | Crawling and issue triage | Debug pages stuck in Discovered or Crawled, currently not indexed |
| Local SEO | Google Business Profile and map rank | Scan a 5×5 Maps grid from your pin and compare with everyone above you |
| Content and Keyword Research | Demand validation and drafting | Build a content plan from your domain alone, with no seed keywords |
| Competitor Research | Backlink and keyword gaps | Referring domains that link to a competitor but not to you, ranked by authority |
The two that are genuinely different
Most of that list is a chat wrapper around things an SEO tool has done for a decade, and that is fine. Two are not.
Fan-out query clustering. When an assistant answers a question it usually runs several searches of its own behind the scenes. Those fan-out searches are where the citations are won, and they are not the same as the question the user typed. Encited pulls those, finds the ones where you are never mentioned, and clusters them into pages to write. Reading the machine’s questions rather than the human’s is the right level to work at, and very few tools expose it.
Off-site citation targets. The same server ranks the domains that AI already cites in your space, including Reddit, LinkedIn and Facebook threads, and points you at them. Being named on a page that already gets cited is a real mechanism, and it is the one that matters most for anyone who cannot outrank the incumbents directly.
Faz says: That second one is the whole game if you are small. Nobody is citing your homepage. They are citing the roundup you are mentioned in, and there is a finite list of those.
What your plan decides, which is the part to read twice
The MCP has no price of its own. It is a way into data your Encited subscription already collects, so the plan you are on sets the ceiling on what any prompt can return. This is the single most useful thing to understand before connecting it, and it is buried in a 43 row comparison table behind a button on the pricing page.
| What bounds the agent | Essential $49 | Business $149 | Agency $349 |
|---|---|---|---|
| Tracked prompts | 15 | 50 | 100 |
| Fan-out queries | No | Yes | Yes |
| Content Agent credits | 50 | 500 | 1,500 |
| SEO Audit pages per audit | 3,000 | 10,000 | 30,000 |
| API access | Render API only | Full analytics and Data API | Full analytics and Data API |
| Extra tracked prompts | From $70/month on every tier | ||
Prices are billed annually. All three tiers include a 14 day trial with no card, and there is a Basic plan at $19 a month covering one site, roughly 10,000 renders and three prompts.
The entry plan cannot run the headline workflow
Look at the fan-out row again. Fan-out queries are not available on Essential.
That matters more than a single missing feature normally would, because fan-out clustering is one of the three sections Encited leads its own MCP page with, and it is the engine behind several of the flagship prompts. Someone who reads the launch page, subscribes at $49 and connects the MCP expecting to cluster fan-out queries into a content plan will find that the data is not on their plan. The workflow the page sells is a $149 workflow.
This is not hidden. It is in their published comparison, which is more than most of this category gives you. It is just three clicks from where the promise is made, and the promise does not carry the caveat.
Credits are the other quiet ceiling
Content Agents run on credits, and the prompts that do the interesting work are the ones that invoke them. Fifty credits on the entry plan against five hundred on the next tier is a tenfold step, and it is the difference between trying the workflows and running them as a habit.
Work out roughly how many agent runs a month your actual routine needs before choosing a tier, because the upgrade trigger here is usage rather than features, and usage creeps.
Connecting it
Two routes, both quick. In the desktop app you add Encited from Claude’s connector directory and sign in when the browser window opens. In Claude Code it is one command, claude mcp add --transport http encited https://encited.com/api/mcp, then /mcp inside a session to authenticate.
One thing the directory listing tells you that the marketing does not
Encited’s own page says to install from the official Claude directory, which is true. What the directory page adds is the category: Encited is listed as a Community connector.
Anthropic’s wording on that page is worth quoting rather than paraphrasing. Community connectors “have undergone automated reviews but aren’t verified by Anthropic”, and it advises connecting only developers you trust.
Read that accurately rather than as a warning. The large majority of directory connectors are Community, it is the normal state for a tool from a company this size, and being listed at all means it passed automated review. But if you are connecting this to a client’s Search Console, the distinction between automatically reviewed and vendor verified is one your own risk process may care about, and it is better to know it now than to be asked in a meeting.
Two questions to settle before you connect it to a client account
Connecting an MCP is an authorisation, not an installation, and the sign-in window is the moment to be awake. Encited’s own flow signs you into your Encited account, which is the account already holding your Search Console connection, so what the agent can reach is whatever that account can reach.
First question: which properties. If one Encited account covers several clients, an agent working on behalf of one of them can see the others, and that is a conversation to have with yourself before it is one you have with a client. Second question: what writes back. Several of the published prompts end in a change rather than a recommendation, and one of them applies edits in your codebase through Cursor. Know which of your prompts are read-only before you let anything run unattended.
Saru says: Neither of those is an Encited problem, it is true of every MCP you will ever connect. It is just that this one reaches your live search data on day one, which is earlier than most.
The prompt library is useful even if you never subscribe
One genuinely unusual thing about this launch: the prompts are readable without an account. Around thirty of them sit on the public page, each one describing a full workflow, which content agent it calls, and what it hands back.
That is worth an hour of anyone’s time regardless of what tool they use. Read as a document rather than as marketing, it is a decent map of what an AI-search workflow actually consists of in 2026: mine a year of Search Console for decay, find the fan-out searches behind AI answers, rank the domains that already get cited in your space, match titles against the queries that earn impressions, and check whether a page is genuinely unindexed or merely uncrawled.
Most of those can be run against whatever data you already have. If you have a Search Console export and an assistant, you can execute a fair few of them today with no Encited account at all, which is an odd thing for a vendor to publish and to their credit they published it anyway.
Faz says: Steal the workflows. That is not cynicism, it is what a public prompt library is for, and the ones worth paying for are the ones that need data you cannot get yourself, which here is the fan-out queries and the citation sources.
Where Encited came from, and why it shows
One observation that explains a lot about the product. Of the 43 features in Encited’s own comparison table, seventeen are pre-rendering infrastructure: edge proxy sites, monthly render volumes, render overage rates, cache refresh frequency, per-path cache rules, content signing, a dedicated render queue.
Encited started as a pre-rendering service, largely for client-side rendered sites built on tools like Lovable, and most of the testimonials on the MCP page are still about exactly that. The AI visibility layer came later, and the MCP sits on top of both.
That history is not a criticism, but it does tell you where the depth is. If your problem is that your JavaScript site is invisible to crawlers, this is a company that has been solving that specific problem for a while. If your problem is purely AI search strategy, you are buying the newer half of the product.
We cover the pre-rendering side separately in our prerender.io alternatives comparison, and Encited also appears in our AI visibility tools guide. This page is only about the MCP.
Who it suits, on the published evidence
| If you are | Then |
|---|---|
| Already an Encited customer | Connect it. It is included, it costs nothing extra, and the prompt library alone is worth the ten minutes |
| Running a client-side rendered site with crawling problems | The pre-rendering half is the older, deeper part of this product, and the MCP is a bonus on top |
| Chasing fan-out queries and AI citations | Budget for Business at $149, not Essential. The data you want is not on the entry plan |
| An agency with many small sites | Check the per-site edge proxy add-on at $15 a site before assuming the tier limits are the real cost |
| Just curious about SEO MCPs | The 14 day trial takes no card, and the prompt library is readable without signing up at all |
What we cannot tell you
We have not run Encited MCP. This is a launch explainer assembled from Encited’s own pages and Claude’s connector directory, not a test, and it deliberately carries no score.
So the questions that actually decide whether a tool like this is any good are all still open here. Whether the fan-out data is accurate. Whether the ranked fix plans are sensible or merely confident. How many credits a real workflow burns. Whether the agent hands back something you would have written yourself, or something you have to unpick. Those are answerable only by running it against a real site for several weeks, and if we get that access we will publish the result with a score attached, including if the answer is unflattering.
Disclosure: Encited is already listed editorially on AI Tools Bakery and has never paid us for placement or for this article. We are covering the launch because an SEO MCP is a genuinely new shape of tool and worth explaining, and this page states what it is built from so you can weigh it accordingly.
Written by
FazFaz is the founder of AIToolsBakery. Some tools here are tested hands on. Others are assessed from vendor documentation and pricing verified on the live page, and every review says which one it is. Sponsors can buy a position in a guide. They cannot buy the score, the criticism, or silence about a better option.
How we test and how we make money →Frequently Asked Questions
