GSC Magic is a hosted Model Context Protocol service that connects AI agents to Google Search Console, Bing Webmaster Tools, Google Analytics 4 and Tag Manager without requiring users to configure a Google Cloud project or run a local connector.
The free plan covers one site, 200 monthly requests, 10 daily calls and 14 days of stored history. Paid tiers are listed at $19 (Pro) and $34 (Agency) per month through December 31, 2026, adding extended retention, optional write actions and multi-client controls across up to 25 sites.
GSC Magic connects AI agents to Google Search Console, Bing Webmaster Tools, Google Analytics 4 and Tag Manager through a hosted Model Context Protocol service. The company advertises 77 tools for retrieving search and analytics data, inspecting URLs and generating reports, with support for Claude, ChatGPT, Cursor and other MCP clients. Users connect to a remote endpoint instead of configuring a Google Cloud project or running their own connector. (GSC Magic)
The immediate choice for site owners is between occasional analysis and ongoing monitoring. The free plan includes one site, 200 requests per month and 14 days of stored history. Paid plans offer retained history beyond Google’s reporting window, optional write actions and agency management features. Access is read-only by default.
This article analyzes the vendor’s published documentation, not independently tested results. The official pages provide no launch timestamp; the setup documentation is marked October 12, 2026, and the remote-MCP guide October 18, 2026.
What the Agent Can Actually Do
MCP gives an AI application a standard way to discover and call external tools. GSC Magic acts as the bridge between that application and the connected search accounts, handling authentication, caching and server-side analysis.
Search Console: clicks, impressions, click-through rate and position by dimensions such as page, query, country and device; URL inspection; sitemap status; and reports on declining pages.
Bing Webmaster Tools: query and page performance, crawl statistics, keyword research and inbound-link information.
GA4: sessions, engagement and key events by landing page, including traffic attributed to AI services.
Tag Manager: inspection of containers, tags and triggers, with read-only access.
The useful part is combining related evidence in one conversation. An agent could compare pages losing Google clicks with Bing performance, then consult GA4 for landing-page engagement. That saves the manual work of exporting several reports and explaining their columns to a chatbot. ()
GSC Magic says it performs analysis on its servers and returns compact results instead of sending thousands of raw rows to the model. Its integration guide specifies compact JSON responses with explicit truncation flags. The company’s claims of faster answers and lower token bills remain vendor claims, but the underlying mechanism is clear: less raw data needs to enter the conversation. (MCP integration guide)
I’d treat that compactness as a practical advantage with a condition attached. A summary helps an agent prioritize work, but a truncated result may be insufficient for a complete audit. When an answer concerns only the largest declines or highest-impression queries, check its date range, filters and truncation status before treating it as exhaustive.
Free, Pro and Agency: What the Price Buys
GSC Magic lists Pro at $19 per month and Agency at $34 per month during its founding-price period, which runs through December 31, 2026. These are vendor-listed promotional prices, so check the current terms before committing. (Plans and pricing)
Plan
Listed monthly price
Documented limits and distinguishing features
Free
$0
One site, 200 monthly requests, 10 calls per day and 14 days of stored history
Pro
$19 founding price
History retained beyond Google’s 16-month window and access to optional write tools
Agency
$34 founding price
Up to 25 sites, five seats, client groups that scope access, and automated monthly client SEO reports
The daily free-plan limit comes from the remote-MCP documentation. The reviewed excerpts do not specify the paid plans’ request quotas, so the table leaves those figures open.
A Request Is Not Necessarily a Conversation
The free allowance needs more attention than the headline price. An agent answering one question may call several tools: one to identify sites, another to retrieve Google performance, another for Bing and another for a report.
As an illustration, a workflow using five counted calls per check would consume 200 calls after 40 checks, before any additional requests. That is arithmetic, not a measured GSC Magic workflow. Actual usage depends on the client, the question and the calls the agent chooses.
The separate 10-call daily cap also limits bursts. An occasional weekly review may fit comfortably; an extended troubleshooting session could hit the daily allowance even with monthly capacity remaining.
My view is that the free plan is best evaluated against a recurring workflow, not site size alone. A small site checked automatically several times a day can generate more calls than a larger site reviewed once a week.
Stored History and Search Console’s Lookback Are Different
The 14-day free-plan figure describes stored history. GSC Magic separately advertises paid retention beyond Google Search Console’s 16-month reporting window. Those figures concern different parts of the data lifecycle. (History comparison)
The reviewed pages do not explicitly describe the free plan as restricting every upstream Search Console query to the previous 14 days. Equally, a paid retention promise should not be taken as a promise to recover data that was already unavailable before connection.
Before paying specifically for historical analysis, ask when collection begins, how much data is initially backfilled and what remains available after a downgrade. For seasonal businesses, those details matter more than the general promise of “longer history.”
Setup Removes the Google Cloud Project
GSC Magic’s hosted setup replaces the Google Cloud project, OAuth client configuration and local runtime that some self-hosted connectors require.
Create an account and authorize Google access. Google sign-in requests Search Console access plus read-only GA4 and Tag Manager access.
Import properties. Select sites within the plan’s allowance. GSC Magic says Bing sites and GA4 properties with matching hosts are linked automatically, though the relevant sources still need connecting.
Add the MCP endpoint to the AI client. The shared address is https://mcp.gscmagic.com/mcp.
Authenticate and ask an initial question. The documentation suggests starting by listing the connected sites.
“No API keys” needs a qualification: Google authentication is managed through OAuth, but connecting Bing requires a Bing Webmaster Tools API key. Some MCP clients also need a GSC Magic access token.
For Claude Code, the documentation supplies this command:
claude mcp add --transport http --scope user gsc-magic https://mcp.gscmagic.com/mcp
The command registers the remote HTTP server in the user’s Claude Code configuration. The documented next step is to open /mcp, select gsc-magic and authenticate. It does not create Google credentials or deploy a local server.
Client requirements can affect the real cost. GSC Magic’s documentation says ChatGPT custom MCP apps require a paid ChatGPT plan and Developer mode; managed workspaces may also need administrator approval. A free GSC Magic account therefore does not guarantee a free end-to-end ChatGPT setup. (Client instructions)
For custom clients, the endpoint uses Streamable HTTP and OAuth 2.1 dynamic client registration. Dashboard-generated bearer tokens provide an alternative, but registration accepts only specified callback origins and loopback addresses. Other hosted clients may need a token or an allow-listed callback. (Connection details)
Read-Only Is the Sensible Starting Point
GSC Magic keeps agent access read-only by default. Paid users can opt into write tools for actions including sitemap submission, IndexNow and Bing URL submission. The company says enabling write access requires a separate Google consent step. GA4 and Tag Manager access remain read-only in the documentation. (Access model, source permissions)
That separation matters because investigating a traffic drop and submitting URLs are different responsibilities. An agent can produce a useful diagnosis without receiving permission to make changes.
I’d keep the connector read-only until there is a specific, repeatable submission workflow worth automating. Paying for a plan that supports writes is no reason to enable them immediately.
The documentation also provides a stronger restriction for token-based connections: marking an access token Read-only makes that token incapable of writing regardless of the plan’s capabilities. Tokens belong in client configuration, never in a chat, and can be revoked immediately. (Access-token documentation)
The service declares MCP safety metadata, including readOnlyHint and destructiveHint, so clients can describe tool behavior. Those labels help the interface communicate risk; account permissions and token restrictions are the controls to check when deciding what an agent can actually do.
The 77-Tool Catalogue Is Not Every Account’s Tool List
GSC Magic computes the available tool list per account. The integration guide says the result depends on:
Subscription plan.
Google permissions granted.
Connected Bing and IndexNow keys.
Whether write mode is enabled.
Client-group restrictions.
A client should call tools/list to discover its actual capabilities instead of assuming all 77 advertised tools are available. This is particularly relevant for agencies, where a narrowly scoped agent is preferable to one that can inspect every client property. (Tool discovery and filtering)
The published pages contain one meaningful inconsistency. The homepage says every source is available on every plan, and the integration guide says a free Google-connected account sees Tag Manager tools. The main documentation’s source table, however, labels Tag Manager access as Pro and Agency. If free Tag Manager access is decisive, confirm it before building a workflow around it. (Homepage, integration guide, documentation)
Which Site Owners Should Pay?
Start free for occasional, single-site analysis. It offers a concrete way to learn whether an agent’s answers are useful enough to justify a subscription. Watch call usage during a normal review and a troubleshooting session.
Consider Pro for continuing records and automated monitoring. Retained history is valuable when you need comparisons that outlast the upstream reporting window. The site also advertises email digests and traffic-drop alerts; confirm their current plan allocation and frequency, since the reviewed excerpts do not fully spell out those entitlements. (Monitoring features)
Consider Agency for access boundaries and client reporting. Its stronger distinction is the combination of sites, seats, client groups and monthly reports. Those controls address a different problem from simply asking more analytics questions: keeping each person and agent focused on the correct client data.
The strongest reason to choose GSC Magic is the setup work it removes. Its usefulness after connection depends on a less glamorous calculation: how many calls a real workflow consumes, which permissions it needs and how long its records must survive.
Frequently Asked Questions
3 questions
1
Does the Free Plan Require a Paid AI Subscription?
Not universally. That depends on the client. GSC Magic’s documentation says ChatGPT custom MCP apps require a paid ChatGPT plan. (Client requirements)
2
Can an Agent Change GA4 or Tag Manager Settings?
The documented GA4 and Tag Manager connections are read-only. Optional paid write tools cover specified search actions, such as sitemap and URL submissions. (Source permissions)
3
Can a Custom Agent Connect?
Yes, if it supports remote MCP over Streamable HTTP. GSC Magic documents OAuth authentication and bearer-token access, with callback restrictions for dynamic registration. (MCP integration)
