XDQ AgentOS · the secure agent layer

The operating layer for AI agents that do real work.

AgentOS plugs agents into your ERP, CRM, HIS and WhatsApp, remembers how your company works, and puts a human gate and an audit trail on every action.

AGENTOS · EXAMPLE RUNLOGGED
09:02:11Enquiry receivedWhatsApp · new lead from a clinic
09:02:13Context pulledCRM history + pricing sheet from company memory
09:02:15Reply draftedAnswer + two slots for a demo
09:02:16Human gateDiscount above policy → sent to sales head
09:04:40Approved & sentCRM updated · meeting booked · run logged
Human gate
Risky steps wait for a person.
Every action logged
Who, what, when — replayable.
Your data stays yours
Deployed in your cloud account.
Swap models freely
Route each task, no rewrite.
How it works

Plugs into what you run. Runs the work. Shows its working.

Connectors read and write to your systems. Company memory gives agents context. The runtime plans and acts — and stops at a human gate whenever a step is risky.

Plugs into
ERPCRMWhatsAppEmailTallySAPHIS · EMRSheetsPaymentsTicketing
XDQ AgentOS
SECURE · YOURS
Connectors
your systems, typed
Company memory
docs, data, history
Agent runtime
plans & acts
Human gates
approve the risky step
Guardrails
policy that can't be skipped
Security scan
find · fix · verify
Audit trail
every run, logged
Model router
swap models, no rewrite
Runs the work
Leads qualified24/7
Claims checkedpre-submit
Invoices reconcileddaily
Tickets resolvedin minutes
Reports writtenauto
Modules

Eight parts. One layer you own.

01
Connectors
Typed, scoped access to ERP, CRM, HIS, email and WhatsApp — read and write.
02
Company memory
Your documents, data and past decisions, retrieved with permissions.
03
Agent runtime
Plans multi-step work, calls tools, retries, and knows when to stop.
04
Human gates
Approval rules by amount, risk or customer — on WhatsApp, email or Slack.
05
Guardrails
Policies the agent cannot override: limits, blocked actions, tone.
06
Security scanning
Prompt-injection and tool-abuse tests run by our security team.
07
Audit trail
Every input, decision and action logged and replayable.
08
Model router
Pick the right model per task; change it later without a rewrite.
Where it runs

One layer, every industry we serve

HealthcareClaim pre-checks before the TPA sees them→FintechKYC triage and follow-ups→MaritimeLaytime and demurrage drafts→LogisticsFreight RFQ to quote in minutes→ManufacturingDealer orders from WhatsApp to ERP→Real estateEnquiry to site visit, same hour→
Why a layer, not a plug-in

AgentOS vs a chatbot bolted on

WHAT MATTERS
Chatbot plug-in
XDQ AgentOS
Access to your systems
Copy-paste or one read-only API
Typed read/write connectors, scoped per agent
Knows your business
Starts from zero each chat
Company memory with permissions
Risky actions
Either blocked or unchecked
Human gate by rule — amount, risk, customer
Proof of what happened
A chat transcript
Full audit trail, replayable
Security
Vendor's promise
Tested and fixed by our security engineers
FAQ

Questions about AgentOS

Is AgentOS a product we buy separately?

No. It is the layer we deploy with every agent build. You get the code, the configuration and the audit trail at handover.

Which AI models does it use?

The model router picks a suitable model per task. You can change models later without rewriting your agents.

Where does our data go?

AgentOS is deployed in your cloud account. Access to each system is scoped per agent and every action is logged.

Can it act without a person?

Only where you allow it. Anything above your thresholds — money, risk or a key customer — waits at a human gate.

Next step

See AgentOS run on one of your workflows.

Send the workflow. You get a scope and a fixed price within two working days — or a straight answer that it isn't a fit.

Secure by default · find → fix → verify · authorised scope only
Send a brief →enquiry@xdqlabs.com