Dangerous actions precede answers
An agent may propose exfiltration, messaging, deletion, payment, or execution before producing a normal-looking final response.
ACTION-LEVEL SECURITY · EVIDENCE BEFORE ENFORCEMENT
Make every AI and critical-system action verifiable before execution.
SenseL turns Agent, Tool, MCP, and existing IT/OT security signals into replayable evidence—so teams can compare versions, validate controls, and retain policy and human approval for high-risk actions.
ONE SENSEL PLATFORM · SECURITY OPERATIONS + AGENT ASSURANCE
Action-Level EvidenceReplayable ValidationHuman ApprovalOn-Premise PathIT · OT · AI Agents
THE ACTION GAP
Risk is not only what the model says. It is what the system actually does.
When AI agents can send mail, query data, change systems, or call MCP tools, security must extend beyond text filtering to multi-step trajectories, authority, actual action outcome, and versioned retest.
Version drift breaks conclusions
A model, prompt, tool schema, MCP server, or permission update can invalidate a one-time security result.
Trace and operating evidence is fragmented
AI, AppSec, SOC, and audit teams lack a shared action trace, decision rationale, and retest record.
Immediate blocking can harm operations
Untested enforcement can interrupt legitimate workflows, making observe-first calibration and human approval essential.
SENSEL AGENT ASSURANCE
Move from a paid assessment to a security release gate that reruns on every version.
Agent Assurance reuses SenseL evidence, validation, on-premise, and human-approval capabilities: Assess finds action-level risk, Validate productizes the evidence, and Gate moves toward pre-action policy through an observe-first path.
01
Available · ScopedPAID SECURITY ASSESSMENT
SenseL Agent Assess
Authorized attack-surface review, testing, remediation guidance, and retest across agents, tools, MCP, permissions, data flows, and guardrails.
Threat modelReplayable traceRemediation + retest02
Design PartnerCORE RECURRING PRODUCT
SenseL Agent Replay / Validate
Turn validated failures into customer-specific regression packs that compare model, prompt, tool-schema, policy, and permission versions.
Version comparisonRegression packCI release gate03
Roadmap · Observe-firstPRE-ACTION POLICY
SenseL Agent Gate
Combine schema, allowlists, identity and permissions, DLP, and trajectory risk before tool execution—observe and calibrate before scoped enforcement.
Schema + allowlistIdentity + DLPHuman approval04
Research OptionOPEN-WEIGHT RESEARCH
SenseL Agent Lens
Explore pre-action internal signals for self-hosted models where weights and activations are accessible; not a prerequisite for closed APIs or the core SLA.
Pre-action signalPer-checkpoint calibrationMatched-baseline gate
SENSEL CLOSED LOOP
One SenseL loop for security operations and agent-action validation.
SenseL brings IT, OT, edge, and Agent/MCP trajectories together with episode evidence, AI reasoning, safe validation, and improvement tracking in one control and validation plane.
01
Collect
EDR, OT, CTI, Agent, Tool, MCP, and energy telemetry
02
Contextualize
Assets, identities, permissions, networks, data flows, and business context
03
Reason
AI correlation, prioritization, and attack narratives
04
Validate
Scenarios, replay, purple team, and version regression testing
05
Preserve
Episodes, policy decisions, human confirmation, and audit history
06
Improve
Remediation, work items, re-test, and baseline updates
Most tools stop at detection, alerts, or prompt classification. SenseL turns signals and agent actions into replayable episodes, then validates whether controls and improvements work within a defined version and scope.
Explore the SenseL PlatformBUILT DIFFERENT
Our moat is not a bigger model. It is a more trustworthy evidence system.
From agent action traces and versioned evidence-contract research to reproducible evaluation, OT edge sensing, and local healthcare AI, SenseL builds trust into the architecture—not the slide deck.
EDGE + OT INTELLIGENCE
Engineering Preview · Public RepoSenseL RelayGuard
Combines active EdgeX telemetry with passive SPAN/TAP sensing at the edge, sending only feature summaries, security events, and evidence references northbound.
Public MVP: OT rules, IEC 61850, MQTT, and Suricata/Snort bridges. SenseL OT Dashboard integration remains on the roadmap.
View public repositoryEVIDENCE CONTRACT
Research Preview · Pre-submissionTrustEpisode
Research into packaging EDR/NDR investigations as versioned, replayable, auditable evidence contracts that keep missing evidence missing instead of letting AI fill the gap.
Public research artifact. It does not imply paper acceptance or full integration of every contract into the production SenseL platform.
View research artifactREPRODUCIBLE EVALUATION
Evaluation Artifact · R0/R1AVLBench
Uses frozen manifests, fixtures, checksums, and separated Safe and Attack-enabled conditions to test whether AI-assisted security validation can be independently recomputed.
Public R0 artifact verification and R1 fixture recomputation are available. R2 live execution remains behind an authorized lab gate.
View reproducible evaluationHEALTHCARE EDGE AI
Research Preview · Private LabEdgeMed
Local-first passive IoMT/OT security research: detectors own the facts, a constrained small language model formats evidence, and deterministic fallback takes over when generation fails.
Advisory only. It has not completed clinical-use, production hospital, or medical-device regulatory validation.
WHERE WE START
Start where proof matters most.
SMB supply chains, critical OT and fabs, and healthcare environments start from different constraints—but all need consistent context, evidence, and safe validation.
SMB SUPPLY-CHAIN SECURITY
01
Build an operational, evidence-ready supply-chain security baseline—fast.
Start with critical assets, accounts, and customer requirements; connect existing or default security signals; and create human-reviewed incident rhythm, remediation backlog, and evidence packages.
AI SOC / AI XDRDefault or existing EDR / NDR integrationWAF and firewall event integrationEpisode Evidence
Explore solutionCRITICAL OT & FAB
02
See the edge and OT environment before validating controls.
Start with passive telemetry, assets, and data lineage, then add protocol-aware events and controlled validation within the operating boundary.
CT / MQTT / Gateway Data FoundationData quality and hierarchy mappingISO 50001 EnB / EnPIM&V and management review material
Explore solutionHEALTHCARE RESILIENCE
03
Validate resilience without disrupting clinical operations.
Connect existing defense signals, move threat hypotheses into isolated or low-disruption workflows, and preserve approval, test, and retest evidence.
Threat Intelligence MappingHealthcare AI XDRWeb / API Security AssessmentIoMT / OT Passive Visibility
Explore solutionFIELD EVIDENCE & DELIVERY PATTERNS
Turn operating constraints into deliverable, retestable security methods.
Each example is labeled according to the evidence currently cleared for publication—as an anonymous delivery pattern, PoC evidence model, or service blueprint—without using unapproved customer identity, deployment scale, or quantified results.
Regional mid-sized hospital — anonymous
Delivery pattern · endpoint scope pending approval
Turn fragmented EDR alerts into traceable operations without disrupting clinical care
SenseL starts with the hospital's existing EDR, connects asset, identity, network, and web context, and turns fragmented signals into human-reviewed Episodes, remediation records, and isolated retest evidence.
Existing EDR integrationHuman-reviewed EpisodesIsolated retest
Review case methodInternational semiconductor manufacturer — anonymous
EnMS + OT PoC evidence model
Use energy CT telemetry to build AI-assisted OT context and improvement evidence
SenseL connects current-transformer telemetry, gateways, equipment hierarchy, and data-quality context to create traceable EnB, EnPI, and M&V workflows. Energy deviations can then be reviewed alongside OT asset and network context, with AI-assisted prioritization and engineering confirmation.
CT / gateway lineageEnMS + OT contextOn-premise · observe-first
Review case methodSMB supply-chain security
Rapid-onboarding service blueprint · not a deployment claim
Turn supply-chain security requirements into an actionable, evidence-ready operating baseline
SenseL starts with critical assets, accounts, data flows, and customer requirements, then connects existing or default EDR, NDR, WAF, and identity signals to create a human-reviewed incident rhythm, prioritized remediation backlog, and deliverable evidence package.
Scoped rapid onboardingControl-to-evidence mappingContinuous improvement cadence
Review case methodINTEGRATE, DON'T REPLACE
Keep existing investments. Add context, evidence, and validation.
SenseL does not require a complete replacement of the security or AI stack. Standardized data, traces, and workflows increase cross-tool operating value.
EDRNDRWAFFirewallSIEM / LogIdentityOT GatewayCT / MQTTCTIAgent TraceMCP / ToolsCI / CD
Keep your stack. Add the intelligence, evidence and validation layer.
TRUST BY DESIGN
High-trust environments need clear data, authority, and evidence boundaries—not AI alone.
Data isolation, least privilege, human approval, and audit trails keep AI-assisted operations and agent tool actions governable and traceable.
Deployment Choice
SaaS, on-premise, or isolated paths are described only where the actual service supports them.
Human Approval
High-risk recommendations and validation actions retain approval, authority, and audit history.
Evidence Preservation
Preserve event sources, AI reasoning, human confirmation, actions, and re-test context.
Data Minimization
Public-site and validation workflows follow clearly bounded, minimum-necessary data principles.
EVIDENCE HUB
We do not only tell you what we are building. We publish evidence you can inspect.
Review public repositories, research artifacts, reproducible evaluations, and safe validation methods—each with an explicit stage.
Open Engineering
Engineering PreviewSenseL RelayGuard
Explore the evidence path from OT edge to the SenseL control plane through EdgeX, passive mirrored traffic, and protocol-aware events.
View public repositoryResearch Artifact
Pre-submissionTrustEpisode
Research into replayable, auditable evidence contracts that preserve missing-data state in AI-assisted EDR/NDR investigation.
View research artifactEvaluation Artifact
R0/R1AVLBench
Use checksums, frozen fixtures, and separated conditions to recompute scoped AI-assisted security validation results.
View reproducible evaluationOT Validation Research
Pre-submissionEviGrid
Turn IEC 61850 threats and CVEs into evidence-grounded, fail-closed controlled validation plans.
View research packageDEFINE A MEASURABLE FIRST PHASE
Do not begin with blanket enforcement. Start with one risk you can replay, remediate, and retest.
Align on the agent or operating scope, existing signals, authority, and acceptance evidence—then define a deliverable, measurable assessment or pilot.