Searchable
Technical SEO, crawlability and public discovery remain the foundation.
A website operating layer for the Agent web
Turn your website into an Agent-ready business card
Enter a public website address. We read the pages a normal visitor can see and create a source-backed business Agent ID that explains who the company is, what it publishes and what an AI system can ask about it.
Patent Pending · US provisional applications 64/122,848 · 64/142,921 · 64/142,927 · 64/142,929 · 64/142,930
- Methodology is public
- Owner confirmation remains explicit
- No source or deployment access
Business Agent ID
Loading featured company cards
One public URL. One machine-readable business profile.
From being visible to being usable.
Answerable
Clear entities, structured content and source evidence help answer systems understand the site.
Recommendable
Fixed-question tests and citation evidence show where the brand appears and where it does not.
Operable
Agent ID, A2A, MCP and accessible interfaces prepare approved tasks for execution.
One identity, two ways to use it
Agents discover it. Owners control it.
An Agent reads the installation contract and prepares a bounded plan. The website owner reviews the exact permissions and approves every write.
Canonical Agent entrypoint
Read the manifest first.
Fetch the target-bound manifest, verify the published release in CLI preflight, create a no-side-effect plan and stop at the owner approval gate.
Production install manifestGET · application/json · no side effect Open
A2A Agent CardAgent ID, interfaces and implemented skillsJSON
Authority modelCredentials, permissions and owner gatesMD
Installation skillCanonical sequence and stop conditionsMD
QuickstartShortest safe path from discovery to verificationMD
Advanced protocol contracts
OpenAPI HTTP operations
MCP Local stdio tools backed by this API
A2A task
Full install guide No-side-effect plan
POST /a2a/v1/message:sendWebsite owner control
Nothing changes before you approve it.
Use the exact HTTPS domain you control. The dry run names every local file, permission and rollback action before installation.
- 1Inspect
Review the release, target domain, files and requested grants.
- 2Approve
Choose the bounded work your own Agent or engineer may perform.
- 3Verify
Publish domain proof and compare the result with an independent public refetch.
Step 1 · Website
Turn a one-time check into a verified website identity.
Load the target-bound production plan for the exact HTTPS website you control. Nothing is written in this step.
Step 2 · Permissions
Review the minimum access before installing.
The list comes from the same production manifest used by the installer. Private keys remain with you, and the mother service receives no source or deployment access.
- Default authority
- Read only
- Website writes
- Owner approval required
- Rollback
- Required for bounded changes
Step 3 · Install
Install the website layer, then connect your application stack.
JavaScript／Node installs the website identity. Python installs the production customer API SDK after the website installation; it does not replace the website installer.
- 1Dry run. No side effect.
Loading command… - 2Owner-approved apply.
Loading command… - 3Public domain verification.
Loading command… - 4Read-only managed inspection.
Loading command…
Production PyPI
Install the customer API SDK from the official PyPI release.
python -m pip install xagent-idUse Python for customer API calls after the JavaScript／Node website layer is installed and verified.
Step 4 · Verify
Confirm the public installation independently.
xAgent ID refetches the website proof from the public network. The browser does not submit a success claim, credential or event.
Public installation status
Run the check after deploying the website layer.
Step 5 · Dashboard
Your verified baseline is ready.
Use the customer-held identity to create a one-time Dashboard pairing URL. The Dashboard starts with the verified website state and the next evidence-backed improvement.
Create a secure one-time Dashboard link
Open Dashboard after pairing
Installation proves technical readiness; it does not guarantee ranking, citation, recommendation or revenue.
Loading command…Behind the card
The identity card is backed by a real installation contract.
The card is the readable result. This is the operational layer underneath it: release, permissions, evidence and rollback remain explicit for owners and Agents.
Verified domainexample.com
xAgent IDIssued at install
Loading
- Package
- From production manifest
- Default access
- Read only
- Allowed writes
- Owner-approved AI files only
- Proof
- Receipt + public refetch + rollback
DiscoverApproveVerifyImprove
Human discoverySearch & AI answersSEO · AEO
Agent discoveryBuyer & service AgentsASO · A2A · Agent ID
Read the live manifest
Trust and control
Useful information should also be accountable information.
Every report points back to a methodology, a schema version and the evidence available at the time of review. Candidate claims remain candidate claims until an owner confirms them.
- 01
Connect
Verify the domain and establish the customer-owned Agent identity.
- 02
Measure
Audit SEO, AEO, GEO and AGO with observed evidence kept separate from missing data.
- 03
Improve
Send a bounded recommendation to the website Agent for local approval and execution.
- 04
Prove
Compare the site receipt with a public refetch and retain a rollback path.
Designed for real customer websites
Your website remains yours.
xAgent ID is not a remote administrator. It operates through public reads, customer-held credentials and explicitly approved AI-related files.
- Identity boundary
- Portable website Agent aliasIt is not Google Agent Identity or a Google principal.
- Default
- Read onlyPublic pages and published Agent contracts.
- Optional write
- BoundedOnly owner-allowlisted AI files, with approval.
- Never requested
- General controlNo broad source repository or deployment access.
- Outcome claims
- Evidence onlyMissing data stays pending. Synthetic outcomes do not count.
Plans
Readable, then measured, then transactable, then governed
Start by making sure an agent can read you and cannot mistake you for someone else. Add measurement when you need to know whether it came. Add a governed channel when agents should be able to reach you. Add bounded execution when every change has to be reversible and provable.
| Capability | Be readable, and unmistakably youIdentityUS$12.99monthly · per verified domain · GACreate Agent ID | RecommendedKnow whether AI came, and what it readMonitorUS$29.99monthly · per verified domain · GAActivate Monitor | Let agents reach you — nothing publishes until you approve itAgenticUS$99monthly · per verified domain · Early AccessActivate Agentic | Change things safely, and prove itOperateUS$199monthly · per verified domain · Enterprise PilotActivate Operate |
|---|---|---|---|---|
| Your official Agent IDOne identity bound to one verified domain, so an agent can tell you from someone claiming to be you | Included | Included | Included | Included |
| A profile a machine can readIdentity, capabilities and sources published in the formats agents actually fetch | Included | Included | Included | Included |
| Installable files generated from your sitellms.txt, Agent Card, schema and WebMCP built from your own public text, with the three fastest fixes; you install them, re-generate any time | Included | Included | Included | Included |
| Proof that AI crawlers cameWhich AI crawlers requested your site, when, and what they got, read from your own CDN logs | — | Included | Included | Included |
| A record of what changedReceipts show what changed on your public surfaces and what was rechecked afterwards | — | Included | Included | Included |
| What to fix first, on the audit you runA scheduled monthly report is not sent yet. Missing evidence stays pending, not zero. | — | Included | Included | Included |
| An inbox agents can useOther agents send scoped, structured requests through a governed channel instead of a contact form | — | — | Included | Included |
| Nothing published without your approvalProposed changes stay pending until a person approves them | — | — | Included | Included |
| Enquiries traced back to the AI answerA sensor on your own pages recognises visitors arriving from an AI assistant and attributes what they do next | — | — | Included | Included |
| Reversible changes with signed proofAllowlisted AI files only, re-verified by an independent public fetch. One-step rollback with a signed receipt, once a change is applied | — | — | — | Included |
| Markup applied in your own edge workerStructured markup is rewritten into the response stream by a worker you run, so no source-code access changes hands | — | — | — | Included |
| Enterprise controls, and no write access to your siteApproval separation and operating guardrails. The parent service never holds credentials to your website | — | — | — | Included |
| Your own governance policyApproval separation, evidence retention and audit export set by you. An inspection priority lane for due work; finishing inside 24 hours is not promised | — | — | — | Included |
Google Cloud
Already building your AI stack on Google Cloud?
Google Agent Identity requires separate Google Cloud setup; binding stays pending until Google evidence is verified. Google Cloud Marketplace purchase is unavailable. Operate checkout is self-serve after domain verification; checkout does not set up Google identity or email execution receipts.
Paid features require a verified domain and confirmed payment. Data sources, inboxes and customer runtimes must be configured for the features that use them; checkout alone does not make them ready. Missing evidence stays unknown or pending. Stripe processes payment details. Subscription changes use Stripe's billing portal; displayed proration applies and cancellation defaults to the end of the paid period.