Ostelun brings discovery, evidence, authorization, and re-verification into one defensible workflow—so security teams can govern AI agents without turning governance into another spreadsheet exercise.
Ostelun separates the questions that enterprise programs too often collapse together: What exists? What proof do we have? Who authorized it? What changed?
Build an AI-agent candidate inventory from bounded control-plane discovery and normalized security signals—with source provenance and exact identity keys retained.
Turn requirements into an evidence plan. Keep machine, hybrid, and human proof distinct so reviewers can see exactly what supports each conclusion.
Make the operating decision explicit. Claiming an agent never means it is approved; authorization is a separate, evidence-backed state.
Track material change against the conditions that supported authorization and selectively re-check what actually changed.
A single lifecycle connects discovery to evidence-backed authorization without erasing the trust boundary between them.
Ostelun connects assurance state to the agent’s identity, tools, permissions, dependencies, and proposed actions—while keeping observed facts, declared baselines, and policy decisions distinguishable.
Compare observed tools and permissions to approved baselines, surface broad privilege and state-changing capability, and leave missing telemetry explicitly not evaluated.
Map Ostelun controls and active Least Agency signals to all ten OWASP Top 10 for Agentic Applications 2026 risk areas.
Project known identities, models, tools, MCP servers, data sources, permissions, connectors, and declared dependencies into a provenance-aware inventory.
Evaluate proposed actions against assurance state, approved tools and permissions, Least Agency findings, impact, and recorded human approval before execution.
Ostelun is being built around conservative identity, provenance, and authorization semantics so a security team can defend not just the answer—but how the answer was reached.
Read Ostelun’s security principlesBound both Ostelun’s collection access and each agent’s approved tools, permissions, autonomy, and human-approval boundaries.
Direct collection, external ingestion, fixture data, and human review remain distinguishable.
Names and fuzzy similarity do not silently merge identities. Contradictory strong keys fail closed.
An Ostelun authorization or runtime policy state is not an enforcement action unless a customer explicitly connects an authorized adapter or gateway.
Give reviewers and stakeholders a compact view of the evidence, decisions, and current proof state supporting an AI agent’s authorization.
The Passport reflects Ostelun’s recorded assurance state. It does not imply certification by a third party.
We’re looking for security, AI governance, and platform teams willing to apply Ostelun to real enterprise AI-agent workflows and help prioritize the controls, integrations, and operating experiences that matter most.