59 MCP tools. 8 slash commands. Your agent runs the SEO.
The Spider ships a local MCP server that gives Claude, Cursor, or any MCP agent hands-on access to your crawls, Google data, and workflows — project-scoped, read and safe-write only, so it acts fast and can never break things.
Claude, Cursor, Windsurf — any MCP clientSafety enforced by the server, not a prompt
What's the worst page hurting acme-store right now?
Used 3 tools
list_issuesget_page_reportget_gsc_top_pages
It's /products/category/running-shoes — high impressions, almost no clicks, and invisible to AI crawlers:
/products/category/running-shoesTop offender
Fixable in one pass
AI-bot accessBlocked
Canonical URLWrong target
ContentThin — 120 words
Fix the canonical and draft unique copy.
Used 2 tools
apply_fixgenerate_content_brief
Done. Canonical fixed and 480 words of unique copy drafted. Re-crawling to confirm.
Reply to Claude…
From prompt to shipped fix, without leaving your editor
Connect once, then talk to your site the way you talk to your codebase. The agent queries real crawl data on your machine — nothing is scraped or guessed.
Connect
Point your agent at digispot-seo
Add the Spider's MCP server to .mcp.json (or Claude Desktop / Cursor settings). It binds to one project — the scope every call is locked to.
Ask
Query anything the Spider knows
Issues ranked by traffic at risk, per-page reports with screenshots, crawl-to-crawl diffs, Google index verdicts — 59 tools cover every module.
Act
Fix with real context
Your agent edits code with the audit evidence in hand, then starts a re-crawl over MCP and compares audits to prove the fix landed.
Automate
Launch whole workflows
One run_workflow call starts a recipe — draft a page, generate an on-brand cover, build a quick-wins spreadsheet — and polls until it's done.
Durable workflows your agent can launch
Recipes are node graphs, not scripts. The Page Writer alone chains six nodes — internal-link targets from your real crawl, a research outline, the MDX draft, an on-brand cover idea, the image render, the finished post — and every completed step is saved, so an interrupted run resumes where it stopped instead of starting over.
Run in parallel
Crawl SiteGet data
url:yoursite.com
render:mobile + desktop
Audit PagesGet data
scope:all pages (auto)
checks:300+
KeywordsData
track:your terms
ranks:(auto)
BacklinksData
target:yoursite.com
new/lost:(auto)
converge ↓
AI InsightsAI synthesis
input:← all four
output:ranked fixes
+:content brief
Built so you can trust an agent with production SEO
Startups and in-house teams don't have time to babysit automation. Agentic SEO only works when the guardrails are structural — enforced by the server, not by prompt engineering.
Project-scoped by design
Every session is locked to one project. A call that targets anything else is refused by the server — your other clients' data is unreachable.
Read + safe-write only
Agents can query everything and run crawls or recipes — but can never delete projects, edit billing, or silently rewrite your business facts.
59 tools across every module
Crawl control, issue analysis, page reports, GSC + GA4 + URL Inspection, keywords, backlinks, knowledge, and workflows — one server, full coverage.
Your key or ours — unmetered AI
Bring your own OpenAI, Claude, Gemini, or DeepSeek key and AI generations aren't metered by us — you pay only your vendor bill. Or use managed credits, no setup.
Durable, resumable runs
Workflow steps are memoized as they complete. Crash, quit, or lose power mid-run — retry resumes from the last completed step, never from zero.
Local-first privacy
The agent talks to the app on your machine. Crawls, reports, and screenshots stay local — only what you explicitly send to an AI provider leaves.
Run it from Claude or Cursor
One prompt, real work
A scripted exchange — but every tool and recipe named here is real and matched by name.
Claude·digispot-seo · MCP
Find this week's quick wins and put them in a spreadsheet I can hand to the team.
Used 2 toolsget_quick_winsrun_workflow
Ran the Quick Wins recipe: 14 fixes ranked by ROI — 3 missing canonicals on money pages lead. CSV is ready; want me to apply the canonical fixes and re-crawl to verify?
Any MCP-capable client: Claude Code, Claude Desktop, Cursor, and Windsurf are the ones we see most. The Spider ships a standard MCP server (digispot-seo) over stdio — if your tool speaks MCP, it works, and the 8 packaged slash commands work anywhere MCP prompts are supported.
Is it safe to let an agent loose on my SEO project?
Yes — by architecture, not by trust. Every session is locked to a single project and the server refuses calls outside it. Agents get read and safe-write access only: they can query data, start/pause/resume crawls, and run workflow recipes, but can never delete projects or data. Knowledge writes are review-first — an agent can propose a business-facts draft, but nothing becomes truth until you approve it in the app. The MCP server is available on paid plans (Tracker and Leader).
Do AI generations cost credits — or can I use my own key?
Both, your choice. Use Digispot managed credits for zero-setup AI, or bring your own key for OpenAI, Anthropic Claude, Google Gemini, or DeepSeek — stored locally, encrypted, and called direct to the vendor so your content never passes through Digispot. With your own key, AI generations aren't metered by us at all: drafts, insights, rewrites, and Image Studio runs cost only your own provider bill, with no per-generation credit. It's the unlimited-AI path for teams that already have an LLM account. (AI features are on paid plans; crawl page-limits and keyword/backlink credits are separate and set by your tier.)
What can my agent actually automate end to end?
The full loop: crawl → query issues ranked by traffic at risk → fix code with the evidence in hand → re-crawl → compare audits to prove the win. Beyond that, one run_workflow call launches any of 11 recipes — from Quick Wins to Spreadsheet to AI - Page Writer (MDX) and AI - Image Studio — and the agent polls the run and downloads the results. Workflow runs are paid-plan features.
Do I need to know SEO to use this?
No — that's what the packaged skills are for. Install the free digispot-ai-seo-skills for Claude Code and type /seo-audit for a graded baseline, then /seo-quick-wins for this week's ranked fixes. Each returns paste-ready remediations, and connecting GSC + GA4 weights everything by your real traffic.