Layer 1
Independent observation
See intent, tool use, process, file, network, and workload activity from outside the agent's control.
AI Governance
AI governance for every supported agent.
Independent observation, security findings, and signed evidence delivered to your existing SIEM.
Run coding agents in a microVM with Brig, free and open source.
Works with the agent tools and execution environments your teams already use.
Capabilities
One system. Three layers of agent governance.
Observe what agents do, turn risky activity into security findings, and preserve the evidence your organization needs.
Layer 2
Security findings
Connect agent context to risky behavior and deliver normalized findings into the SOC.
Layer 3
Signed governance evidence
Retain tamper-evident records with integrity status, evidence source, and declared coverage.
Evidence model
One evidence model across every supported workload.
Connect supported developer machines, runners, automation, and customer-facing agents to a record built outside the agent boundary.
Supported workload evidence
Independent observation and signed records
Coding-agent sessionBoundary and host activity
Self-hosted runnerJob and process activity
Internal automationSigned evidence
Customer-facing agentFinding delivery to SIEM
Evidence model
Signed evidence with declared coverage
CoverageDeclared
SourceBoundary + host
DeliveryExisting SIEM
Who it is built for
AI governance for every stakeholder.
CISO
Approve autonomy with evidence.
Review what agents did, where the evidence came from, and what remains outside coverage.
Platform teams
Connect workflows without replacing them.
Add independent observation to supported machines, runners, and agent environments while keeping the surrounding stack.
Developers
Move fast. Keep the workflow.
Use supported agents and sandboxed execution while governance runs outside the agent-authored record.
Why Probity
Probity records agent activity outside the agent.
Swipe across to compare every evidence field.
| Outside agent boundary | Intent + execution | Signed evidence | SIEM delivery | Local + remote | |
|---|---|---|---|---|---|
Probity Independent observation, findings, and evidence in one model. | |||||
Agent-native logs Useful session context produced inside the agent environment. | |||||
Endpoint-only telemetry Host activity without the agent's full objective and delegated context. | |||||
Sandbox-only controls Isolation without a complete investigation and governance record. | |||||
SIEM-only ingestion A destination for events, not the source of independent agent evidence. |
One record. Every supported agent workload.
Independent evidence flows into the security stack your organization already operates.
Built by open-source systems creators
Built by the creators of urunc.
Probity is powered by NOFire AI and built by the creators of urunc, an Apache-2.0 cloud-native runtime that is now a CNCF Sandbox project.
Open-source lineage
- Project
- urunc
- Purpose
- Open cloud-native runtime
- Foundation
- CNCF Sandbox
- License
- Apache 2.0
Common questions.
Join the Probity waitlist.
Get AI Governance updates. Brig is free and open source and available now.