Skip to main content
Your AI app picks the right tool for your question, so you rarely need to name one. This page shows what is available, so you know what to ask for.
The nine manage tools change things in Spyro AI. They appear only when your key or sign-in allows Read and take actions, and every change waits for you to confirm it. See Confirming changes. Everything else works with Read.

At a glance

Read your own data (free)

Research and ideas (10 credits per call)

Make changes (Read and take actions)

The manage tools themselves cost nothing. Some actions spend credits, just as they do in the dashboard. See Credits and limits.

Your site

getWorkspaceSnapshot returns your business profile, competitors, SEO and GEO scores, audit issues, keywords and recent content. It shows only the workspace you connected.searchMySite takes a query (2 to 200 characters) and k (1 to 10 results). It searches your indexed pages and returns the URL, title, heading and an excerpt.Try: “Which of my pages already cover cold brew coffee?”
getAuditIssues is the quick list. Optional filters: severity (critical, warning, info), category (seo, geo) and limit (1 to 25). Returns each issue with the page, a plain-English message and a suggested fix, most severe first.getSiteAudit reads the whole audit. It takes a view:
  • overview (default): scores, status, issue counts, blocked AI crawlers.
  • issues: one group per issue type. Pass type for that type’s pages. Add withFixes: true to get ready-to-paste code (file, where to put it, language) for robots.txt AI crawler rules, canonical tags, the viewport tag, Open Graph tags, Organization schema and a draft llms.txt.
  • pages, page (one page with its issues and score history), history and trackedUrls.
Try: “Show my critical audit issues with fixes, then apply them in this repo.”
getRankSummary returns your top tracked keywords by position and the five biggest improvements.getKeywords takes a view: searches (saved keyword searches), keywords (the rows of a search, with questions or gems filters), analysis (a keyword your organization already analyzed), tracked, rankHistory and rankSummary.Try: “Which keywords moved up the most this week?”
Takes a view: settings (every workspace setting, the brand and the writing voice), account (plan, trial, credits left, articles and keyword searches left this month, prompt and keyword slots) or integrations (Search Console, Bing, GA4 and publishing destinations).Try: “How many credits and articles do I have left this month?”

Search data

getSearchPerformance (Google Search Console) takes a view: overview, live, queries, countries, sitemaps, lowCtr, almostRanking, decay, trend, branded, appearance, discover, indexing, inspectUrl. Other options: window (28d or 90d), sortBy (clicks or impressions), limit (1 to 50), days (7, 28, 90 for trend), and url (required for inspectUrl). URL inspection is limited to 20 per workspace per day. If Search Console is not connected, you get a clear “not connected” answer. getBingPerformance takes a view: overview, queries, pages, crawl, crawlIssues, links, sitemaps, keyword. Search Console and Bing data is read live and cached for about three hours. Try: “Which queries have lots of impressions but a low click rate? Suggest new titles.” or “Which pages are losing traffic?”

AI visibility

getAiVisibility takes a view:
  • citations: which prompts cite you, the sources AI engines use, and which competitors are cited instead.
  • responses: the raw answers from ChatGPT, Gemini, Claude, Perplexity and Google AI Overviews, with sentiment.
  • aiOverviews: whether Google shows an AI Overview for your tracked keywords and whether you are cited.
getVisibilityDiagnosis returns a ranked list of actions based on your audit and tracked prompts. getPrompts goes deeper. Its views: list (every prompt with its citation rate over the last 4 checks), prompt (one prompt’s trend, results per engine and rivals), answers, mentions (every answer of the last 30 days), fanout (the searches the engines ran while answering), search, suggestions and gapAnalysis. analyzeCitedSources (10 credits) takes 1 to 5 urls that AI engines cite and explains why they win, with content actions you can copy. Try: “Where am I losing AI citations to competitors, and what should I publish to fix it?”

Research

For keyword volume, difficulty and Google results, use manageKeywords with research or analyze, then read the result with getKeywords. Try: “Run keyword research for espresso machines, then show me the low-difficulty questions.”

Content and articles

recommendBlogIdeas (10 credits) and planContentCalendar (10 credits) propose ideas. They never write to your calendar. count is 1 to 10, and topic and seedKeywords steer the ideas. getArticle lists your articles (20 a page, newest first, with status, keyword, word count and SEO score). With article (its id or exact title) it returns one article with its SEO fields and full body as markdown or html. With format: "mdx" a finished article comes back as one .mdx file with frontmatter and JSON-LD, ready to save in your site’s repo.

Making changes

Each manage tool takes an action and the fields that action needs. Name an existing item by its id or its exact title or text. If two items match, the tool returns the candidates so the app can pick one by id.
add (title and primary keyword, date optional), update (title, brief, writer options, listed websites), move (to a date, swapping with any idea already there), delete, sendToWriter (writes the article now, 250 credits) and findWebsites (searches again for the websites a list article features). New ideas are checked against Google as they are added.
edit (replace the Markdown), setKeyword, setSeo (meta title, meta description, slug, focus keywords), delete, retry (a failed article, 250 credits when it finishes), skipImages (open an article whose images are stuck) and rewrite (one passage, 10 credits). rewrite returns the new text without saving it; save it with edit.
publish, schedule (an ISO 8601 time up to 365 days ahead), cancelSchedule, pause and resume (workspace admins only). Name the destination by its id or name. Free. Connecting a destination stays in the dashboard.
add, pause, resume, setFrequency (daily or weekly), delete, runNow (once per 24 hours), retry (engines that failed), suggest (up to 10 new prompts, free, 3 runs a day), trackSuggested, draftFromSearches (article ideas from the searches AI engines ran), analyzeGap (why answers cite other sites, free, 20 a day) and gapToCalendar. Each check costs 10 credits per engine, 20 for Claude.
research (10 credits and one of your monthly keyword searches), deleteSearch, analyze (20 credits the first time your organization analyzes that keyword in a market, then free for 30 days), track (each rank check costs 2 credits) and untrack.
run, addPages (up to 10 pages), markFixed, markNotFixed, trackUrl, recheckUrl and untrackUrl. All free.
startScan (up to 5 competitors in one country, 100 credits per competitor fetched, free when fetched in the last 30 days), refreshOwnData (free, 3 a day), deleteCompetitor, writeArticle (an article for a keyword gap), track and dismiss.
update (any workspace setting; read the current values with getSettings first), recrawl, regenerateBrand, regenerateTone and resync (gsc, bing or ga4, once an hour each). Free. Billing, team, keys, connecting integrations and deleting the workspace stay in the dashboard.

Confirming changes

Nothing changes until you confirm it in Spyro AI:
  • Apps that can ask you in the chat show a Spyro AI confirmation with a summary of the change. Confirm there.
  • Other apps get a link like https://spyro.app/mcp/confirm/.... Open it, check the summary and click Confirm. Then ask the app to repeat the same call, and it returns the result.
A link works once, only for the person who connected the app, and expires after 10 minutes. If the app calls again while the link is still open, it gets the same link back. The summary names any credits the change will spend.
Read the summary before you confirm. An AI app can misread your request, and some actions (delete, publish) cannot be undone.

Prompts for coding agents

Apps that support MCP prompts get three ready-made prompts. In Claude Code they appear as slash commands named after your server, such as /mcp__spyro-ai__fix-audit. They use only free read tools.

Working with several workspaces

When you sign in and connect more than one workspace, the app gets one extra free tool, listWorkspaces, and every tool accepts a workspace argument (the workspace id from that list). The app should call listWorkspaces first. Keys connect one workspace only, so they never need this.

Example prompts

Weekly SEO review

“Check my audit issues, rank changes and Search Console for the last 28 days. Give me the 5 most important things to do this week.”

Find content gaps

“Show my competitor gaps, find topics they cover that I don’t, and check my content inventory so you don’t repeat anything.”

Plan a month

“Recommend 8 blog ideas for next month, check each against my existing pages, then add them to my calendar on Tuesdays and Thursdays.”

Win AI citations

“Run a visibility diagnosis, show where competitors are cited instead of me, and audit my top page for what to change.”