Agentimus
Agentimus helps AI assistants like ChatGPT, Claude and Perplexity find your site, read it correctly, and cite it in your own words — and shows you which AI bots are actually visiting. You don’t need to understand AI or web standards to use it: a setup wizard walks you through everything in about a minute on your first visit, then it runs on its own.
Want more control? You also get a first-party log of every AI crawler that fetches your content, one-click blocking for the bots you don’t want, and a one-screen readiness report that scores how AI-ready your site is — and tells you the next thing to improve.
By default it makes no outbound requests, collects no analytics, and logs no IP addresses — everything runs on your own site. The optional AI Visibility feature is the one exception: turn it on and add your own AI provider API key, and it queries that provider to check whether AIs cite you (see External services).
📖 Full documentation — a plain-English user manual and a developer reference, with step-by-step guides for every feature: https://heera.github.io/agentimus/
Control — who may use your content
- robots.txt content-signals + AI-training blocklist — declare your content-usage policy and block named model-training crawlers (GPTBot, CCBot, ClaudeBot, Google-Extended, Bytespider, …) by name, while leaving read/cite bots free.
- Block scanners & scrapers (opt-in hard block) — robots rules are a polite request; this enforces them. Turn it on to return 403 to the user-agents on your denylist, and optionally auto-deny agents that disguise themselves as ancient handsets (a classic scanner trick). Your always-allowed list is never blocked — pre-trust well-known AI assistants and answer engines (ChatGPT, Claude, Perplexity, …) with one click, while the major search engines (Googlebot, Bingbot, …) are recognised and trusted automatically.
/.well-known/acme-challenge/(SSL renewal) always stays reachable.
Reduce exposure — what your site reveals to bots
- Exposure controls (opt-in, all OFF by default) — a panel of switches that quietly close the things stock WordPress reveals to anonymous crawlers and scanners: stop username enumeration (the
?author=1and REST/wp/v2/usersleak, plus the users sitemap and oEmbed author), 404 author-archive pages, hide the WordPress version from the generator tag and asset URLs, drop the rarely-used auto-generated<head>discovery links, and neutralise XML-RPC. Nothing changes until you turn a switch on, and signed-in admins and the block editor are never affected. It’s exposure hygiene, not a firewall — Agentimus stays a discovery layer, not a security suite.
Visibility — who is reading you
- Agent activity log — a dashboard of which AI crawlers and agents actually fetch your content and endpoints (GPTBot, Claude, Perplexity, Googlebot, …), recorded first-party in your own database, with no IP logging.
- Activity to review — a nav-bar queue surfaces the clients worth a second look — new, unusually high-volume, or spoofing what they are — names a recognised crawler where it can, and offers one-click Block or Allow (trust). Nothing is blocked unless you choose to.
- AI Visibility (opt-in) — track each brand, product or person you choose across ChatGPT, Perplexity, Gemini and Claude. For every one, Agentimus asks the questions your audience actually types and reports whether it gets mentioned, linked, and how it ranks against its own rivals — over time. Each thing you track has its own website, competitors, questions and scoreboard; pause any single one, or the whole schedule, whenever you like. Off by default; you bring your own API key for each engine, and this is the one feature that makes an outbound request (see External services).
Content — clean, machine-readable output
- Markdown delivery — request any page as clean markdown by appending
.mdto its URL (or, where your server allows it, with anAccept: text/markdownheader). - /llms.txt & /llms-full.txt — an llmstxt.org index of your pages, topics and recent posts, plus a full-text edition an agent can ingest in a single request.
- JSON-LD — WebSite + Person/Organization, plus BlogPosting and BreadcrumbList on posts. Automatically defers to Yoast, Rank Math, SEOPress, AIOSEO and The SEO Framework so you never ship duplicate schema.
- Topics for AI — say what each post is about in plain words, right in the editor; those topics become the JSON-LD
keywordsand a line in the page’s.md, so assistants understand each page’s subject. Type your own, or let Agentimus fill them in from the post’s own tags and categories. Nothing shows on the visible page. - XML sitemap — an opt-in fallback sitemap (index + paginated sub-sitemaps), generated only when neither WordPress core nor an SEO plugin already provides one, and advertised in robots.txt and llms.txt.
- Change feed — a JSON feed at
/agentimus-changes.jsonlists your recently added, updated and removed pages, with a?since=filter, so an assistant re-checks only what changed instead of re-reading your whole site. On by default and advertised in your discovery document.
Identity & contact
- Author / site identity — a profile sentence, expertise topics and linked profiles (
sameAs) feed llms.txt and JSON-LD — the highest-signal lines for agent retrieval. - security.txt — optionally publish an RFC 9116 disclosure contact at
/.well-known/security.txt, so researchers and agents have a machine-readable way to report an issue.
Readiness report
- A one-screen score of how machine-readable your site is, with a plain-English checklist of what’s enabled and what’s still missing.
- Agent preview — open it from the Readiness tab to see the exact JSON-LD and Markdown an AI agent receives for the whole site or any page or post, then copy it. It shows what would ship even when the feature is off or an SEO plugin owns your schema, and a matching read-only preview also sits right in the post editor — so you never have to view page source to check what agents read.
- AI Readability tips — as you write, an “AI Readability” panel in the post editor flags what makes a page hard for an assistant to read and cite: thin content, missing headings, no opening summary, a nav-heavy page, or images without alt text. It sits in the same “Agentimus” box as the per-page Agent preview, so you check what an agent receives and how readable it is in one place. Editor-only — nothing shows to visitors.
Machine discovery (forward-looking)
Agentimus also publishes a single, normalized discovery document, built to the conventions the agent ecosystem is converging on (the .well-known convention, A2A agent cards, MCP-shaped tools). It puts a site’s identity, capabilities and APIs in one predictable place:
- /.well-known/discovery.json — an owner-curated document describing the site’s identity, capabilities, APIs and agent cards. Other plugins can declare themselves through a single optional hook, so what an agent needs is aggregated in one place.
- /.well-known/agent-card.json and /.well-known/mcp.json — an A2A agent card and an MCP manifest, generated automatically.
- Standards-aligned
.well-knownendpoints — an RFC 9727api-catalog, plus — only when the capability actually exists — an MCP server card and an Agent Skills index. Optional response signing (Web Bot Auth / HTTP Message Signatures, RFC 9421): sign the discovery documents with an Ed25519 key published at/.well-known/http-message-signatures-directory, so agents can verify they came from you. On by default; the private key stays on your server. - WordPress Abilities API → MCP tools — registered abilities are projected into MCP-shaped tool descriptors, and a running MCP server (if one is installed) is detected and linked. Agentimus advertises tools; it does not execute them.
- Zero-config auto-discovery — reads your registered REST API namespaces, public post types and the WordPress Abilities API, so a site is described even when no plugin declares itself. A Discovery Hub admin screen shows what an agent can see, and you decide what is published.
What’s read today vs. what it readies you for
Honest framing: the content signals above (JSON-LD, robots, llms.txt, markdown) are read by search engines and AI tools today. The discovery document is forward-looking and standards-aligned — it prepares your site for AI agents as they adopt these conventions, rather than claiming every agent already reads it. The discovery format is an open, openly-licensed convention with a public reference, not a private one, and the plugin works fully whether or not anything consumes that document.
Why it’s useful
Most tools cover one slice — an llms.txt file, an AI-bot blocker, or structured data. Agentimus brings content control, agent-traffic visibility, clean machine-readable output and a forward-looking discovery document together in one coherent, lightweight package — and tells you what’s still missing.
AI readiness is also called AI SEO, GEO (Generative Engine Optimization) and AEO (Answer Engine Optimization) — publishing the machine-readable signals AI systems need to find, read and correctly represent your site.
External services
By default, Agentimus does not connect to or send any data to any external service: it makes no outbound HTTP requests, loads no remote scripts, fonts or analytics, and stores the agent-activity log in your own database with no IP addresses.
The optional AI Visibility feature is the only part that calls an external service, and it is off by default. When you enable it and add your own API key for one or more AI providers, Agentimus sends the prompts you configured to those providers to check whether they mention and cite your site. This happens only for the engines you turn on, and only when a check runs — either when you click “Run check now” or on the schedule you set. Your API keys are stored on your own site and are used solely to make these calls; nothing else is sent anywhere. The providers you can enable — and their terms and privacy policies — are:
- OpenAI (ChatGPT) — https://openai.com/policies/terms-of-use · https://openai.com/policies/privacy-policy
- Perplexity — https://www.perplexity.ai/hub/legal/terms-of-service · https://www.perplexity.ai/hub/legal/privacy-policy
- Google (Gemini) — https://ai.google.dev/gemini-api/terms · https://policies.google.com/privacy
- Anthropic (Claude) — https://www.anthropic.com/legal/consumer-terms · https://www.anthropic.com/legal/privacy
The generated discovery documents contain a $schema value that names the document format (in the same way a schema.org URL identifies a vocabulary). It is a label inside the output only — it is never fetched.
The example URLs in examples/integrate-your-plugin.php (on example.com) are placeholders for documentation; they are not requested by the plugin.
Source & build
There is no minified-only code. The admin interface is built from Vue 3 source in resources/ with Vite; the source and vite.config.js ship in this package and also live in the public repository at https://github.com/heera/agentimus . Run npm install && npm run build to regenerate assets/admin/ from source.