Not marketing copy — the real registry, the real grounding rules, the real orchestration engine, and the real open standard it's built on. Every number on this page is live.
Every one of the tools on this platform runs through the same real code path — there's no per-tool bespoke backend hiding behind each of the 354 cards on the catalog page. What actually differs between "CV Score Analysis" and "Concrete Mix Advisor" is which real data grounds the answer, not which engine answers it.
activityRegistry.js) of every real tool — id, label, domain, icon. No tool exists that isn't a real key in this file; the API 404s on anything else rather than guessing.This is the actual switch statement (services/featureGrounding.js) every one of the 354 tools passes through before the model sees the request. It is deliberately honest about which domains get live data versus a static reference anchor versus nothing — a tool with no live data source still gets a genuine AI answer, it just isn't tagged as verified.
| Domain | Grounding source | What it actually does |
|---|---|---|
| Salary Intelligence | ● live data | Queries a real salary-benchmark collection built from real market data, not a model guess. |
| Intelligent Job Hunt | ● live data | Queries real, currently-indexed job postings for the role/location asked about. |
| Career Goals | ● personal record | Pulls this specific user's own real saved goals — never another user's, never fabricated ones. |
| LifePath AI | ● live data | Builds context from real, already-computed career-trajectory data for that role/country. |
| Civil Engineering | ● named standards | Anchors on real, named codes (ACI 318, AISC 360, Eurocode, AASHTO) and forces a licensed-engineer sign-off line on every calculation-shaped answer. |
| Architecture | ● named standards | Anchors on real named frameworks (RIBA Plan of Work, LEED/BREEAM) and flags anything jurisdiction-specific as needing local verification. |
| Everything else (CV, Cover Letters, Interview, LinkedIn, Brain AI, Tools, Enterprise) | ○ no live source | A genuine AI answer with no external anchor to check it against — same honesty level as any other AI chat, just not claiming to be verified. |
[VERIFIED] for an established fact, [INFERRED] for reasoned extrapolation, [ESTIMATED] for a calibrated guess, [UNKNOWN] when it genuinely doesn't know — and when real fetched data appears above, that exact figure gets tagged [LIVE DATA] instead.The Workspace Agent is the one place the "354 separate tools" framing breaks down on purpose — it plans a real sequence of them, chains each real output into the next, and only ever executes a step after either the system validates it against the real tool list or a human explicitly approves it.
Every real AI call — a single tool or one step of a Workspace Agent run — is tried against a real fallback chain, not a single provider with no plan B. This exists because a single-provider platform goes fully dark the moment that one provider has an outage; the chain exists to keep answering anyway.
The Model Context Protocol is Anthropic's real open standard (donated to the Linux Foundation's Agentic AI Foundation in December 2025) for connecting an AI model to external tools and data without a bespoke integration per pair. This platform implements a real MCP client, not just the word "MCP" in a marketing sentence.
@apify/actors-mcp-server, wired to two real actors: a Google Places crawler and a website-content crawler. A real, valid Apify API credential is already configured in this environment.Live from the same registry the API serves — this is documentation, not the interactive version. To actually run one, use the catalog page.
Real citations, not invented statistics — anything not a first-party fact about this platform's own code links to where it actually came from.