Votre SEO et vos analyses — directement dans votre assistant IA

List the websites in your Astrina account (domain, title, SEO score). Call this FIRST to discover which domains the other tools can query.

One-call snapshot of a site: visitors & pageviews, top traffic sources and countries, SEO score, latest audit grade, tracked-keyword visibility and backlinks.

Privacy-first web analytics for a site: visitors, pageviews, visits, and breakdowns by source, country or page — or a daily time series.

The latest full SEO audit for a site — overall score, pass/warn/fail counts and the prioritised list of issues to fix. Set refresh:true (or if none exists) to queue a fresh 141-check audit.

Keyword rankings for a site: current search position, weekly change, landing URL and clicks for each tracked keyword, plus an overall visibility score.

Backlink profile summary for your account (total / live / lost links and status breakdown).

Internal business metrics a site pushes to Astrina (users, registrations, materials, payments, moderation …) with today's value and 30-day change.

List the Astrina SEO/web utilities runnable over the API (name + what each does), so you know what to pass to astrina_run_utility.

Run one Astrina utility on a URL or value — e.g. headers, meta tags, robots.txt, SSL, redirects, DNS. Use astrina_list_utilities for the names.

Live status of the Pulsar distributed device network: devices live/enrolled, verified checks in the last 24h, consensus rate and a 0-100 health score. No arguments.

Check whether a URL or domain appears in the Pulsar threat feed (phishing / malware / scam) before trusting or clicking it. Returns listed + category + sources.

The Pulsar network's honest, measured AI benchmark: how often the AI agrees with human consensus on held-out golden tasks, human calibration accuracy, and total labels resolved. Values are null when there is not enough data. No arguments.

Where the Pulsar network is right now: the top countries by verified checks in the last 7 days. No arguments.

How much of the Pulsar fleet can run AI on-device: AI-capable devices, live devices, fleet coverage percent, and verified on-device inferences. No arguments.

Fetch a PUBLIC web page from the Pulsar network vantage — a stable egress IP that gets past many bot-walls that block AI agents. Returns final URL, HTTP status, title, description, readable text (truncated), content SHA-256 + SimHash, latency and the vantage IP. Use for grounding/RAG when your own fetch is blocked or you need a neutral vantage.

PROOF-OF-OBSERVATION: fetch a public URL and return a cryptographically SIGNED statement (Ed25519) attesting to exactly what existed at that URL right now — content hash, SimHash, title, timestamp. Anyone can verify the signature against the network public key (pulsar_pubkey) offline, forever. Use to create tamper-evident evidence, provenance, or a citation you can prove later.

Verify a proof-of-observation from pulsar_observe: checks the Ed25519 signature over the canonical statement against the network public key. Pure verification, no fetching. Returns valid true/false + the statement.

Ground a factual claim against the live web: fetch the URL from the Pulsar vantage and check whether the claim is actually supported by the page. Returns a verdict (supported / likely_supported / partial / not_found), keyword overlap, an evidence snippet, and a SIGNED observation of the page so you can prove what it said. Use before asserting a fact from a source (anti-hallucination).

Has a page changed? Fetch a URL and return its current content SHA-256 and SimHash. Pass a hash you saw before to get changed true/false and a SimHash distance — so you know whether your cached/RAG copy is stale.

Deterministic, verifiable content fingerprint (32-bit SimHash) of text or a fetched URL — the same locality-sensitive hash every Pulsar device computes, so near-duplicate content gets near-identical hashes (compare by Hamming distance). Cheap dedup/similarity for RAG. Provide "text" OR "url".

The Pulsar network Ed25519 public key (hex) used to sign proofs of observation. Use it to verify pulsar_observe signatures offline. No arguments.

Verify an AI agent actually completed a web action: fetch the URL and confirm the expected text/result is present now — with a signed observation as proof. Use after your agent posts/edits/publishes to confirm it went live.

Geo vantage: measure a URL from a specific COUNTRY on the Pulsar device network (reachability + speed as seen from there), or list the countries currently available. Call with no country to see available vantages.

How AI sees a brand/domain: ask a live LLM what it knows about a domain and whether it recommends it, and detect whether the domain is mentioned. Useful for AEO / AI-visibility monitoring. Returns the answer, mentioned true/false, and the model used.

Build a small provenance-stamped dataset from public URLs: fetch each, dedupe near-duplicates by SimHash consensus, and return a manifest with a deterministic dataset hash + a signed attestation. For RAG/training corpora whose provenance you can prove.

Embed text into a vector using the network inference layer: a real neural embedding when a model is configured, else the deterministic 32-bit SimHash. Returns the vector (or its dimensions + hash) and which backend produced it.