I like the big SEO suites. Ahrefs and Semrush are excellent when you need to explore. What I did not want was five people answering the same Monday questions in five different tabs:
- Did anything material move?
- Do we already have a fresh enough snapshot for this query?
- What should a human — or an agent — do next?
So we built seo-studio: a small, opinionated SEO ops workbench. Not a public “OpenSEO for everyone.” Not a pixel clone of anyone’s UI. Just our surface for buy → store → review → analyze, with the same jobs exposed over MCP.
This is a build note — architecture, cost shape, and habits. It is not a ranking playbook and not a teardown of proprietary product features.
The problem we were actually solving
Full suites optimize for discovery: click around, open twenty reports, follow a hunch. That is valuable.
Our weekly ops questions were narrower and more repetitive:
- Did we already pay for this SERP / keyword payload?
- Is the stored copy still fresh enough for this decision?
- If we analyze it with an agent, can we prove where every number came from?
Without a named store, teams re-fetch the same queries, paste screenshots into chats, and eventually invent “approximate” metrics when the API is slow. We wanted the opposite culture: if it is not in the store with provenance, it does not go in the report.
Published cost shape (why we bothered)
List prices change; treat these as published entry points (Sep 2026), not our invoice:
| Option | Published entry | Seats / model | Best when |
|---|---|---|---|
| Ahrefs Lite | $129 / mo | 1 user; extra seats billed separately (often cited ~$40/seat on Lite) | Daily deep research in a polished UI |
| Semrush Pro | $139.95 / mo | 1 user; add-on seats | Same — exploration UX |
| DataForSEO SERP | ~$0.60 / 1K Standard queue (~$1.20 Priority, ~$2 Live) | Pay-as-you-go; typically $50 min deposit | You store + analyze yourself |
| seo-studio pattern | API spend + engineering | Shared evidence store + MCP | Repeat ops questions + agents |
Sources: Ahrefs pricing, Semrush plans, DataForSEO SERP API.
Illustrative scenario (not a claim about our bill): three people who occasionally need the same SERP answers. Three suite seats quickly land in the ~$200–$230+/mo neighborhood before you have a shared cache. Five thousand DataForSEO Standard SERPs is on the order of ~$3 in API spend — plus the engineering time to own seo-studio.
Suites still win if people live in exploratory UI all day. Metered + store wins when the pain is repeat questions and agent access, not browsing.
What seo-studio is (and is not)
Is: a workbench for SEO ops evidence — snapshots, review queues, analysis handoff, MCP tools.
Is not: a multi-tenant SaaS, a Semrush clone, or a content autopublisher driven by keyword lists.
We keep schemas, internal scoring, and competitive playbooks out of public posts on purpose. The reusable idea is the loop, not our private tables.
Four layers
1) DataForSEO — metered fetch, not a second brain
DataForSEO is the paid pipe for SERP / keyword / related-style payloads when the knowledge base does not already have them.
Habits that keep the bill sane:
- Prefer a stored snapshot over a new purchase when it is still fresh enough for the decision.
- If the API did not return it, it does not appear in a report.
- Keep request shapes boring and repeatable so cost is predictable.
We are not trying to redraw every suite graph. We pull what the weekly loop needs and stop.
2) seo-studio — the workbench
Browse what we stored. Refresh only when policy allows. Queue reviews. Hand off to analysis.
Intentionally thin UI:
- Honest empty states (“no snapshot yet”) instead of fake confidence.
- Provenance visible next to figures (what we asked, when, which market).
- Exact collection schemas stay internal.
Think ops console, not marketing site feature.
3) Jev — analysis on evidence, not vibes
Jev is our SEO/GEO analysis agent. In this loop it consumes what seo-studio already holds. Only then may it recommend buying more data.
Contract:
- No fabricated metrics.
- Prefer “insufficient evidence” over a confident guess.
- Output should be actionable for humans and for other bots — not a wallpaper of charts.
Calling it an “AI SEO tool” sells the wrong idea. It is closer to a careful analyst with a boring database.
4) MCP — same surface for agents
This is the part that changed day-to-day work: seo-studio speaks MCP.
The same jobs the dashboard can do — check whether we already own a snapshot, request an export, ask Jev for analysis on stored evidence — are available as tools to agents in our workflow.
Why it matters:
- One policy for “do we buy again?” whether a human clicks or an agent asks.
- Fewer one-off scripts that quietly re-hit the API.
- Safer defaults: agents inherit “no invent, show provenance,” instead of each bot improvising a client.
MCP here is not a growth hack. It is how humans and agents share one evidence store.
A Monday loop that stays cheap(ish)
- A question lands (trend, gap, cannibalization, freshness, …).
- seo-studio looks for an existing snapshot.
- If missing or stale for that decision → buy from DataForSEO → write back with provenance.
- Jev drafts findings from stored evidence (or says evidence is insufficient).
- A human picks the content or technical work worth doing.
The boring win is step 2. Most SEO API burn we have seen elsewhere comes from re-fetching the same query because nobody owned a named cache.
Failure modes we design against
- Pretty dashboard, silent guesses. Charts without provenance are worse than no charts.
- Agent bypass. If MCP is incomplete, someone will curl the API “just this once.” Then forever.
- Autopublish from keyword lists. That is how you ship spam and burn trust (on DEV.to and everywhere else).
- Clone the suite UI. You will lose; buy the suite for exploration and keep seo-studio for ops.
Trade-offs (honest)
Pros
- Cost control for repeat, low-breadth questions.
- One evidence store for humans and agents.
- MCP encodes policy once.
Cons
- You own freshness policy, empty states, and migrations.
- You will not get every fancy visualization on day one.
- If your team needs broad exploratory research daily, keep a full suite — use seo-studio as glue, not as a religion.
- Engineering time is real. “API is cheap” does not mean “system is free.”
If you copy the pattern
You do not need our names. The reusable idea is:
Metered SEO API → ops store with provenance → analyst (human or agent) that refuses to invent → optional MCP so agents reuse the same store.
Whether the store lives in your CMS plugin, a small admin app, or plain Postgres is secondary. The failure mode to avoid is a chatty agent (or a pretty UI) fed by silent guesses.
Closing
We built seo-studio because our Monday questions were ops questions, not “open twenty reports” questions. Suites still matter for exploration. DataForSEO matters for rows. The workbench matters so we do not pay twice — in money or in invented metrics.
We run this for our own ops at CoworkingView. Questions about the architecture welcome in the comments.
Top comments (0)