Agents watch & propose
AI agents monitor your connected sources, surface findings and prepare draft fixes.
Trust & security
gx360 is built around one rule: agents and experts can prepare the work, but nothing that touches your live site or accounts happens without your approval.
The access model
Every change follows the same path — prepared by an agent, approved by you, carried out by the agent, and logged.
AI agents monitor your connected sources, surface findings and prepare draft fixes.
Anything that touches your live site or accounts waits for your yes. We ask by email, WhatsApp or a call, and record who approved, how and when.
The agents make the approved change on the connected account and log it — or hand your developer an exact brief. Nothing goes out unapproved, ever.
What changed, when, why and who approved it stays in an audit trail.
We read through Google’s API with the access you grant. We don’t copy your accounts out, and you can revoke access any time.
The optional MCP connector reads your data only — it cannot change anything, and nothing is connected until you do it.
Agents and specialists can prepare anything, but nothing public ships until you approve it.
Data handling
You grant access connector by connector. Here is what gx360 reads, where it lives and how long we keep it.
| Connector | What gx360 accesses | Where it’s stored | Retention |
|---|---|---|---|
| Google Search Console | Read-only: search queries, impressions, clicks and index coverage. | Stays in your Google account; read via Google’s API. | Report snapshots kept for the engagement; removed on request. |
| Google Analytics (GA4) | Read-only: traffic, conversions and channel performance. | Stays in your Google account; read via Google’s API. | Report snapshots kept for the engagement; removed on request. |
| Google Ads | Read, plus the changes you approve: search terms, keywords, conversion goals. | Stays in your Google account; approved changes applied by an admin on your team. | Change log retained; account data is not copied out. |
| Website / CMS | Public pages we crawl; CMS edit access only when publishing is in scope. | Crawled content stored with your gx360 workspace. | Kept for the engagement; removed on request. |
| CRM / pipeline (optional) | Read-only deal and lead stages, only when pipeline tracking is agreed. | Aggregated into your workspace reports. | Kept for the engagement; removed on request. |
| AI model providers | Only the text needed for a specific task, e.g. an audit prompt or a draft. | Processed by the provider to return a result; not used to train their models. | Not retained for training; transient processing only. |
Model providers
gx360 uses large language models to draft and analyse. Today that means OpenAI and DeepSeek, chosen per task for quality and cost. Only the text needed for a task is sent, and your data is not used to train any provider’s models.
Customers with stricter requirements can request an OpenAI-only configuration, so no task is routed to any other provider. When you connect Claude, it reads your gx360 data — read-only — under your Anthropic account.
Start with the evidence
Start the free trial with read-only access, or talk to sales and we’ll walk through exactly what gx360 can and can’t touch.
Want IOSS experts to do the work? Talk to sales