What does the Agent do that Core doesn't?

Core measures traffic, AI visibility, and bot activity. The Agent reads those signals and produces a daily prioritised brief, deterministic recommendations, and Claude-ready prompts. Core is the eyes; the Agent is the queue.

Is the Agent autonomous? Does it deploy code on its own?

No. The Agent helps prepare and control website improvements with Claude. It does not deploy changes without operator approval. Every execution step requires explicit operator action — the Agent generates the prompt; you paste it into Claude; you review the diff; you deploy manually; you click Verify deployment.

Does the Agent guarantee SEO rankings?

No. No tool can. The Agent surfaces operational recommendations based on the data it observes; ranking outcomes depend on many factors outside any single tool's control.

How does the MCP connection to Claude work?

The Agent exposes a Model Context Protocol surface that Claude can connect to. Claude can read your recommendations, repo mappings, and site context through MCP; it can prepare prompts and code changes; the operator reviews and approves before anything is deployed. No autonomous deploy.

What data does the Agent use?

WebmasterID Core events, bot visits, AI referrals, geo signals, page metadata, crawler-derived data, internal-link structure, repository mappings, execution history, and (where supported) imported operational datasets. Everything is workspace-scoped.

What is a repository mapping?

An operator-supplied binding between a site and the repository that owns its code (org/repo + framework + deployment + branch). Mappings are never derived from the domain. Once verified, the Agent includes the repo + likely editable surfaces in the Claude prompt.

Which plan includes the Agent?

Pro, Agency, and Business. The Free plan is WebmasterID Core Lite only. See /pricing for the full breakdown.

Agent

Recommendations, Claude workflows, operator-approved execution

The WebmasterID Agent reads your Core analytics, prioritises the work, and produces Claude-ready prompts you can act on through Claude Code or the MCP connection. Operator approval is required for every change — the Agent never deploys on its own.

Start with Pro See pricing

What it is

A review-first execution layer on top of Core

The Agent is not a separate product — it is the layer of WebmasterID that turns analytics into prioritised work. Daily brief, deterministic recommendations, Claude prompts, repository mappings, full execution lifecycle. Bundled with Pro and above.

Data sources

What the Agent actually reads

Everything the Agent recommends is grounded in data WebmasterID already observed for your workspace. Nothing is invented; empty buckets stay empty.

· WebmasterID Core events (page_view, interactions, server-to-server)
· Bot visits (AI crawlers, search bots, automation tools)
· AI referral attribution (ChatGPT, Claude, Perplexity, Gemini, AI Overviews)
· Geo signals (privacy-safe country / region; no raw IPs)
· Page metadata (URL, pathname, referrer category, traffic category)
· Crawler-derived data (per-site, per-pathname hit history)
· Internal-link structure (where available)
· Recommendation history + memory writeback
· Repository mappings (operator-supplied)
· Execution lifecycle history (planned → executed → verified → re-audited)
· Imported operational datasets (limited today — see /agent-data-import)

Data import

Imported datasets enrich recommendations

The Agent can use operator-supplied data alongside Core. Import support is expanding — the full picture lives on /agent-data-import.

Crawled page data (where supported)
Sitemap data + canonical history
Repository mapping data (org / repo / framework / branch)
Execution history (planned / executed / verified)
CSV / manual operational data, where the format is supported

Imported data is workspace-scoped and feeds the recommendation and task-generation pipelines. Read the full import guide →

MCP

How Claude connects through MCP

The Agent exposes a Model Context Protocol surface. Claude can subscribe to it and read your workspace's recommendations, site context, and repository mappings — through the same allow-listed surface the dashboard uses.

Claude reads recommendations the operator has surfaced.
Claude inspects repository mappings, framework, and inferred editable surfaces.
Claude generates prompts and prepares code changes.
The operator reviews every diff and deploys manually — there is no autonomous deploy and no auto-merge.

Long-form MCP positioning + connection guide: /agent-mcp.

Claude workflows

What Claude does with Agent data

A Claude prompt produced by the Agent carries the org/repo, default branch, the verifyBefore checklist, the rollback notes, and the recommendation lineage. Claude consumes that context and prepares a concrete code change.

Claude reads the recommendation + the prepared context.
Claude proposes a code change (a diff, not a deploy).
The operator reviews the diff against the verifyBefore list.
The operator deploys manually.
The operator clicks Verify deployment to close the loop.

Execution lifecycle

Detect → Prioritise → Prepare → Execute → Verify → Remember

One deterministic loop. Every transition is operator-triggered; every step is audited.

Detect
Core ingests events, bot visits, and AI referrals. Diagnostics surfaces tracker health, geo enrichment, and pipeline staleness in real time.
Prioritise
The Agent dedupes, scores, and groups recommendations into work packages. The daily brief surfaces the top 10 actionable items — never more.
Prepare
One click freezes a snapshot of the prompt, repo mapping, verifyBefore checklist, and rollback notes. The Claude prompt arrives with org/repo, default branch, and inferred editable surfaces.
Execute
You paste the prompt into Claude, review the diff, and deploy manually. The Agent never executes code or git operations on its own.
Verify
Click Verify deployment once the change is live. The Agent records the operator's confirmation triple (deployed + verificationPassed + notes) — no fake automation.
Remember
Memory writeback updates the rule brain: positive outcomes raise confidence; rollback signals flag regressions. Re-audit shows whether the issue is still appearing.

Outcomes

What the Agent helps you improve

The Agent helps prepare operational improvements grounded in your Core data. It does not promise rankings or guarantee any specific outcome.

· Indexability + crawler reachability
· AI-search surface area (which AI engines read your site)
· Page-level metadata + heading structure
· Internal-link coherence
· llms.txt policy and AI-crawler compliance
· Tracking + ingestion health

Safety

Operator approval is required for every change

The Agent helps prepare and control website improvements with Claude. It does not deploy changes without operator approval.

No auto-merge, no auto-deploy, no autonomous git operations.
Every transition writes to admin_audit_logs with actor + action + workspace + entity.
Recommendations + prompts are snapshot-pinned at prepare time; the operator sees exactly what Claude saw.
When the data does not support a recommendation, the bucket stays empty — never fabricated.

Pricing

The Agent is included on every paid plan

Free is WebmasterID Core Lite only — analytics + AI crawler visibility. The Agent is included automatically on Pro, Agency, and Business. CashWorkspace is bundled with every paid tier.

See /pricing for the full breakdown.

Operator stack

WebmasterID is part of a small operator ecosystem

None of these are required to use WebmasterID Agent. Listed here because they compose well with the operator workflow.

Pocket Manager also on Google Play.

FAQ

Frequently asked

What does the Agent do that Core doesn't?: Core measures traffic, AI visibility, and bot activity. The Agent reads those signals and produces a daily prioritised brief, deterministic recommendations, and Claude-ready prompts. Core is the eyes; the Agent is the queue.
Is the Agent autonomous? Does it deploy code on its own?: No. The Agent helps prepare and control website improvements with Claude. It does not deploy changes without operator approval. Every execution step requires explicit operator action — the Agent generates the prompt; you paste it into Claude; you review the diff; you deploy manually; you click Verify deployment.
Does the Agent guarantee SEO rankings?: No. No tool can. The Agent surfaces operational recommendations based on the data it observes; ranking outcomes depend on many factors outside any single tool's control.
How does the MCP connection to Claude work?: The Agent exposes a Model Context Protocol surface that Claude can connect to. Claude can read your recommendations, repo mappings, and site context through MCP; it can prepare prompts and code changes; the operator reviews and approves before anything is deployed. No autonomous deploy.
What data does the Agent use?: WebmasterID Core events, bot visits, AI referrals, geo signals, page metadata, crawler-derived data, internal-link structure, repository mappings, execution history, and (where supported) imported operational datasets. Everything is workspace-scoped.
What is a repository mapping?: An operator-supplied binding between a site and the repository that owns its code (org/repo + framework + deployment + branch). Mappings are never derived from the domain. Once verified, the Agent includes the repo + likely editable surfaces in the Claude prompt.
Which plan includes the Agent?: Pro, Agency, and Business. The Free plan is WebmasterID Core Lite only. See /pricing for the full breakdown.