apify-insights
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@apify-insights Which of my Apify actors lose money?"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
💬 What you can ask
Ask | You get |
"How did I do yesterday?" | Profit against the same days last month, plus the 3 actors to look at |
"Which actors lose money?" | Each one with the price where it would break even |
"Is anything broken?" | Actors that fail, that Apify flagged, or that charge nothing |
"Which actors had failed runs this week, and on which days?" | Failed, timed-out and aborted runs per day and today so far; for one actor, the failed runs users shared, their last log lines and the build before it started |
"Tell me about maps-scraper" | Users, runs, price, builds, reviews and 3 months of money |
"Which actors are leaking users?" | Each actor marked Leaking, Rising, Fading or Converting, with the reason and what to do |
"What should I fix for quality?" | Apify's quality suggestions across your actors, ranked by impact |
"Where do I rank for tiktok scraper?" | Your Store search position and who ranks above you |
"Compare me with another creator" | Your public portfolio next to theirs |
Related MCP server: Apify Store API MCP Server
⚡ Install
Let your AI agent do it. Paste this into Claude Code, Codex, Cursor, Gemini CLI or OpenCode:
Set up apify-insights for me by following https://github.com/Devansh-365/apify-insights-mcp/blob/main/skills/setup-apify-insights/SKILL.mdOr run two commands yourself (needs Node.js 20+):
npx -y apify-cli login # once: connects your Apify account
npx -y github:Devansh-365/apify-insights-mcp setup # adds it to every AI client you haveRestart your AI client and ask "How did my Apify actors do yesterday?"
Revenue and profit need a small Chrome extension. Claude Desktop, one-click installs for single clients, updates and fixes are all in docs/install.md.
💡 Good to know
Read-only. It never changes anything in your Apify account.
Runs on your computer. It talks only to Apify, and your token stays in the Apify CLI login.
No guessing. A number it cannot get shows as "not available", never as zero.
Every tool and how each metric is worked out: docs/tools.md.
🛠️ Development
npm test # tests on made-up data
npm run build # bundles src/ into dist/index.js (committed, so rebuild before you commit)
npm run check:dist # fails if dist/index.js is out of date
npm run pack:mcpb # builds the Claude Desktop extensionWhy it installs this way: ADR 0001. Release notes: CHANGELOG.md.
📄 License
MIT, by Devansh Tiwari.
Available Tools
12 toolsacquisition_scoreAcquisition triageARead-onlyIdempotent
Which actors need attention for acquisition, and what to do: each of your top actors (by 30-day users) is Leaking, Rising, Fading, Converting, Steady or Quiet, judged against the same point last month and against your other actors, with the reason. Rising and Fading are not judged before day 8 of the month. Pass actor for one actor. Same numbers and field names as growth; for raw counts, paying users, referrers or countries use growth.
Inputs: top, actor, month.
actor_profitabilityMoney per actorARead-onlyIdempotent
Per-actor revenue, cost, profit, runs, paying/free users, cost and revenue per 1,000 results, price and break-even price for a calendar month. Views: top, losers, movers (change in revenue per day vs the month before), no_revenue, below_cost.
Inputs: top, view, actors, period, +1 more.
actor_qualityQuality scores and suggestionsARead-onlyIdempotent
Apify's quality score for your actors and its improvement suggestions, grouped by work type (security, operations, schema, pricing, listing) and ranked by impact (work-type weight × traffic). Without actor: the top issues across your top actors by 30-day users, each with the actors it affects. With actor: that actor's score, sub-scores, praises and suggestions.
Inputs: top, actor.
actor_reportActor reportARead-onlyIdempotent
Everything known about one actor: users, 30-day run results, flags, current price, default memory and timeout, latest build, reviews, local source folder, and (with a Console session) the last three months of money with unit costs and break-even price.
Inputs: actor (required).
auth_statusConnection statusARead-onlyIdempotent
Whether the Apify API token works (and for which account) and the state of the Console session used for money data. Call this when another tool returns an auth error.
Inputs: no named parameters.
competitor_benchmarkCompare with another creatorBRead-onlyIdempotent
Your public portfolio vs another creator's: actors, users (30 days), median users per actor, actors with 10+ and 100+ users, total runs, and each side's top 5 actors.
Inputs: username (required).
daily_briefMorning checkARead-onlyIdempotent
Yesterday and month-to-date money (paying users) vs the same days last month, plus up to three actors to look at, each with the rule that flagged it and the next call to make. Falls back to health only without a Console session.
Inputs: no named parameters.
growthViews to runs to paying usersARead-onlyIdempotent
For a calendar month: how many users viewed your actors, ran them and paid, per actor (most paying users, most used, or worst view-to-run conversion). One actor adds top referrers and countries. For which actors need attention and what to do, use acquisition_score.
Inputs: by, top, actor, month.
list_actorsList my actorsARead-onlyIdempotent
The connected account's actors with users (7/30 days), users' runs and failure rate (30 days), last run and any Apify notice. Default: top 25 by 30-day users. Use query to find an actor by name.
Inputs: top, sort, query, visibility, +1 more.
revenue_summaryRevenue summaryARead-onlyIdempotent
Account revenue, cost, profit and margin for any period (yesterday, last7d, a month, a date range), optionally compared with the previous period or the same days last month.
Inputs: period, audience, compareTo, includeDaily.
run_healthRun healthARead-onlyIdempotent
Finds actors that need attention: flagged (Apify notice such as under maintenance, hidden from Store search), failing (3+ failed or timed-out users' runs and 5%+ failure rate in 30 days, aborts excluded) and, with a Console session, no_revenue plus revenue at risk. Returns only problems by default, worst first. Pass actor to check one actor. Pass period (needs the Console session) for byDay: failed, timed-out and aborted runs per UTC day, today included; with actor also the failed runs users shared with you, their log tails, recent builds and the build before the failures began.
Inputs: top, logs, actor, period, +1 more.
store_rankStore rank by keywordBRead-onlyIdempotent
Your actors' positions in Apify Store search for up to 10 keywords, in the order a Store visitor sees, with the actors ranked above your best one.
Inputs: keywords (required), depth.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 12 tools
Most tools target a distinct analysis axis (money, health, quality, acquisition, search rank, competition), and descriptions explicitly disambiguate the risky pairs, e.g. growth vs acquisition_score ('same numbers and field names as growth; for raw counts... use growth') and revenue_summary vs daily_brief. There is still a cluster of four tools answering 'which actors need attention' (run_health, actor_quality, acquisition_score, daily_brief) that an agent could plausibly misselect, though the rule sets differ enough to be workable.
All names are snake_case and readable, but conventions differ: verb_noun (list_actors), noun_noun (auth_status, revenue_summary, run_health, actor_report), and bare noun phrases (growth, store_rank, competitor_benchmark). The actor_ prefix is used for some actor-scoped tools but not others (actor_profitability vs growth), so the namespace is inconsistent.
Twelve tools is squarely in the well-scoped range, and each maps to a coherent analytics question (one actor, all actors, money, health, quality, rank, competitor) rather than redundant variants. Nothing feels padded.
The analytics domain is well covered end-to-end: auth, portfolio listing, per-actor deep dive, revenue/cost/margin, run health, quality, acquisition, store rank, and benchmarking, spanning both overview and drill-down. Minor gaps remain around longitudinal trending/export of metrics and explicit remediation or alerting actions, but agents can work around these via period parameters and date ranges.
Maintenance
Related MCP Connectors
Ask about your Agentio campaigns, Creator deals, and performance on YouTube and Meta. Read-only.
Read-only access for your AI to Search Console, GA4, Google Ads, Meta Ads and your store's sales.
Read-only Google Search Console, Bing, Google Analytics and privacy-friendly analytics for your AI.
Read-only AI search visibility data: citations, AEO audits, and advisor insights from AI-Advisors.
Related MCP Servers
- AlicenseNot gradedqualityDmaintenanceProvides AI agents with real-time web capabilities including live search results, markdown web scraping, business lead generation, and detailed company information. It enables agents to bypass knowledge cutoffs by accessing current web data through a monetized Apify Actor.MIT
- FlicenseNot gradedqualityFmaintenanceEnables querying the Apify Store to retrieve structured data on public Actors including pricing, usage trends, ratings, and categories. Integrates with AI agents for market research and competitor tracking.-
- AlicenseNot gradedqualityCmaintenanceProvides read-only access to self-hosted cash-flow forecasts, balances, transactions, and credit-card data, letting AI clients answer spending and upcoming-obligation questions without modifying financial settings.MIT
- AlicenseNot gradedqualityCmaintenanceEnables agents to access 30 pay-per-event Apify actors for ad, video, reel, and audio transcripts; Google Trends and keyword demand; job boards; Amazon product data; and Instagram data, charging only per delivered item.MIT