Diagnose the real environment
Agents gather evidence through typed connectors instead of guessing from the ticket description alone.
The autonomous IT workforce
Specialized IT agents, from request to resolution.
Opsentry handles help desk and support across the technology stack. Its growing agent library diagnoses problems, executes approved actions, coordinates across domains, documents every step, and resolves tickets end to end.
Tron the front door
Lex the dispatcher
ServiceNow the audit trail
Agents the product
Built to work across the systems your IT team already governs
ServiceNowMicrosoft GraphEntraExchangeDefenderIntuneSlackTeamsVPNSAPNetworkEnterprise appsServiceNowMicrosoft GraphEntraExchangeDefenderIntuneSlackTeamsVPNSAPNetworkEnterprise apps
The operating principle
Autonomous resolution. Never autonomous guessing.
Routine, low-risk work completes without human involvement. Risky work pauses for approval. Ambiguous or unsupported work reaches the right human team with the investigation already complete.
Act inside boundaries
Controlled runbooks expose only registered actions with structured inputs, permissions, and approval rules.
Prove the outcome
Every remediation must be verified. Failed checks, exceptions, and unsupported paths escalate with evidence.
ServiceNow system of record
The ticket holds the whole story.
Product architecture
Four roles, designed around the ticket.
Tron is the front door. Lex is the dispatcher. ServiceNow is the audit trail. The specialized agents are the product—and they own the actual diagnosis and remediation.
The core product
Specialized agents
A growing workforce for every IT domain. Each agent has a narrow identity and operating scope. It can investigate the real customer environment, use explicitly registered actions, coordinate with other agents, document its work, and own a ticket through verified resolution.- Diagnose
- Act
- Coordinate
- Document
- Verify
- Resolve
The front door
Tron
Meet employees where work already happens. Tron lives in Slack or Microsoft Teams. It understands the request, decides whether a ServiceNow ticket is necessary, creates one when appropriate, and reports the verified result back to the employee.- Slack and Teams
- Reasoned ticket creation
- Final result delivery
The dispatcher
Lex
Triage and orchestration, not an all-in-one engineer. Lex normalizes and classifies each ticket, selects the correct agent and runbook, coordinates handoffs, requests human approval when required, and tracks the case to a verified outcome.- Agent selection
- Runbook routing
- Durable case state
The audit trail
ServiceNow
One system of record for every path in. Requests from Tron and tickets created directly in ServiceNow enter the same automation system. Investigations, actions, approvals, handoffs, notes, and resolutions remain attached to the ticket.- Work notes
- Approval tasks
- Verified closure
The operating contract
Agent package
Identity, rules, runbooks, and exact actions. Agents are defined as controlled software packages, not open-ended chatbots. Their capabilities and boundaries are explicit, reviewable, and extensible.- agent.yaml — identity and scope
- instructions.md — rules and safety
- runbooks/*.yaml — deterministic workflows
- Typed actions — exact permitted operations
The agent library
One specialized identity for every part of the stack.
Each agent owns a narrow domain, a declared set of intents and actions, and the runbooks required to carry work through resolution.
Entra Agent
Diagnoses identity, sign-in, account recovery, authentication method, group access, and directory issues through Microsoft Graph and Entra controls.
Exchange Agent
Handles mailbox access, delegates, forwarding, shared mailboxes, mail flow, and other governed Exchange Online support workflows.
Microsoft Security Agent
Investigates security alerts, risky activity, protection state, and approved remediation across Microsoft security systems.
Intune Agent
Diagnoses device enrollment, compliance, configuration, application deployment, and endpoint management issues.
VPN Agent
Investigates connectivity, client configuration, access policy, authentication, and network-path issues before applying an approved fix.
SAP Agent
Owns SAP-specific application health, access, and support checks, then hands the ticket off when evidence points to another domain.
Network Agent
Diagnoses DNS, routing, reachability, wireless, and infrastructure issues within declared connector and approval boundaries.
Generic Software Agent
Supports approved enterprise and end-user applications with product-specific diagnostics, runbooks, and registered actions.
End-to-end workflow
Every request advances toward a verified outcome.
Two ways in. One automation system.
A request starts with Tron in Slack or Teams, or arrives as a ServiceNow ticket. Tron creates a ticket when tracked work is required, and both paths converge on the same durable case.
Request receivedTicket created if neededContext preserved
Lex dispatches the right expert and runbook.
ServiceNow sends the case to Lex. Lex normalizes the request, classifies the domain and intent, then selects the specialized agent and deterministic runbook that can safely own the work.
NormalizeClassifySelect agentStart runbook
The agent works against the real environment.
Using typed connectors, the selected agent gathers evidence from the customer environment and records its findings and internal notes on the ServiceNow ticket before attempting remediation.
ConnectGather evidenceWrite work notes
Act, pause when required, verify, and close.
The agent executes safe, permitted actions. Higher-risk work creates a ServiceNow approval task and resumes from the saved checkpoint after approval. The agent verifies the result, closes the ticket, and Tron reports back.
Act or request approvalResumeVerifyClose and report
Cross-agent coordination
One ticket can move across multiple experts.
An agent can return a case to Lex when the evidence points outside its domain. Lex reassigns the same ticket, preserving every note, action, and diagnostic result.
→
The SAP Agent confirms SAP is healthy. Lex passes the same case and evidence to the VPN Agent, which diagnoses the network path, requests approval, applies the fix, verifies connectivity, and closes the ticket.
SAP Agent confirms the application is healthy and records the evidence
Lex Reassigns the same ticket with the diagnostic context intact
VPN Agent remediates, verifies connectivity, and closes the case
Controlled autonomy
Resolve what is safe. Pause or escalate what is not.
Registered actions only
An agent can invoke only the typed operations explicitly registered for its identity, domain, and runbook.
Structured permissions
Every action has defined inputs, target permissions, approval rules, verification requirements, and audit output.
Approval before risk
Higher-risk actions pause the workflow and create a ServiceNow approval task before any change is made.
Durable checkpoints
Paused work resumes from its saved checkpoint after approval or input instead of restarting or losing case context.
Verification before closure
A successful tool call is not enough. The runbook must gather post-action evidence that the remediation worked.
Explicit human escalation
Ambiguous cases, unsupported systems, protected resources, and failed verification go to the right human team with diagnostics complete.
A continuously expanding workforce
Grow the agent library without rebuilding the platform.
Add a domain, not another platform
Supporting a new IT domain means defining an agent, its runbooks, its typed actions, and its approval policies. The orchestration platform stays the same.
Explore agentsThe Polsia model, applied to IT
Polsia aims to operate a company autonomously. Cofounder is an autonomous company-building partner. Opsentry brings that model to IT help desk and support.
Explore architectureAcross the technology stack
The agent workforce can grow across identity, devices, applications, email, security, networking, infrastructure, and enterprise software.
Explore ecosystemFAQ
How the autonomous workforce fits together.
What is the core Opsentry product?
The product is a growing library of specialized IT agents. Each agent can diagnose a narrow class of problems, execute approved actions, coordinate with other agents, document its work, and resolve tickets end to end.
Is Opsentry one general-purpose IT GPT?
No. Agents are not open-ended chatbots, and Lex is not an all-in-one engineer. Agents reason inside controlled runbooks and can use only explicitly registered, typed actions.
What does Tron do?
Tron is the employee-facing interface in Slack or Microsoft Teams. It understands the problem, decides whether a ServiceNow ticket is necessary, creates one when appropriate, and reports the final result.
What does Lex do?
Lex is the triage and orchestration layer. It classifies the ticket, chooses the agent and runbook, coordinates handoffs, requests approval, and tracks the case until the outcome is verified.
Does Opsentry replace ServiceNow?
No. ServiceNow remains the system of record and audit layer. Requests from Tron and tickets created directly in ServiceNow enter the same automation system, and every meaningful step is written back to the ticket.
When does a human get involved?
Routine, low-risk work can complete autonomously. Higher-risk actions pause for approval. Ambiguity, unsupported systems, protected resources, and failed verification route to the appropriate human team with the diagnostic work already completed.
Build your autonomous IT workforce