Platform Components
Velatir is built around seven components that work together to monitor AI usage, evaluate compliance, and enable human oversight.Organisations & Workspaces
Organisations are the top-level container. Workspaces sit within an organisation and separate monitoring by team, department, or project.
Traces & Sessions
Every AI interaction is captured as a trace. Related traces are grouped into sessions that represent a complete conversation or workflow.
Agents
Agents evaluate every trace for compliance risks across data protection, access control, HR, brand safety, and business risk.
Insights
Agents produce insights into your AI usage and actionable insights that tell you what to do about it.
Review Tasks
When an agent escalates a trace, a review task is created for a human to approve, reject, or request changes.
Channels
Notifications and decisions flow through Slack, Microsoft Teams, Email, or the web dashboard.
Integrations
Data sources like the browser extension, SDKs, n8n node, and MCP server feed traces into the platform.
How Components Work Together
Trace Captured
A data source sends an AI interaction to Velatir as a trace. Traces have a direction: Inlet (incoming to an AI system), Response (outgoing from an AI system), or Signal (a passive event). Related traces are grouped into a session.
Agents Evaluate
All active agents assess the trace in parallel. Each agent focuses on a different domain and independently determines whether to allow, block, or escalate the trace.
Action Resolved
The platform resolves the agents’ assessments based on their roles. An Observer’s finding is logged but takes no action. An Enforcer can block the trace outright or require approval through a review task.
Human Reviews (If Escalated)
If an agent escalates, a review task is created and sent to connected channels. The reviewer sees the trace, the agent’s assessment, and makes a decision.
Key Principles
Traces Are the Foundation
Every AI interaction is captured as a trace. This gives your organisation full visibility into AI usage.Agents Handle the Volume
Agents evaluate every trace automatically across different compliance domains. Humans only review what agents escalate.Roles Control Agent Authority
Each agent operates as an Observer or Enforcer. This lets you start with monitoring only and increase enforcement as you build confidence.Human Oversight Is Built In
Review tasks ensure a human is always in the loop for high-risk interactions. Decisions happen in the tools your team already uses.Everything Is Auditable
Every trace, agent assessment, human decision, and final outcome is logged with full context. Your compliance trail is always complete.Organisations & Workspaces
| Concept | Description |
|---|---|
| Organisation | Top-level container for your company. Manages billing, org-wide settings, and agent defaults. |
| Workspace | A subdivision within your organisation for a team, department, or project. Has its own members, traces, and agent configuration. |
| Org Roles | Administrator (full access) and Reader (view only) |
| Workspace Roles | Admin, Editor, Moderator, and Reader, each with progressively fewer permissions |
Agent Modes per Workspace
Each workspace can configure how it uses organisation-level agents.| Mode | Behaviour |
|---|---|
| Inherited | Uses the organisation’s default agent configuration |
| Configurable | Workspace sets its own agent roles and settings |
| Disabled | Agent is turned off for this workspace |
Trace Directions
| Direction | Description | Example |
|---|---|---|
| Inlet | Incoming request to an AI system | A user submits a prompt to a chatbot |
| Response | Outgoing response from an AI system | The chatbot’s reply to the user |
| Signal | Passive event or notification | A tool call or system log entry |
Integration Options
Velatir captures traces through multiple data sources depending on your setup.- Browser Extension. Monitors AI interactions across 4000+ platforms with no code changes.
- Local Detection. Scans devices for AI tool usage.
- TypeScript SDK. Programmatic integration for JavaScript and TypeScript applications.
- Python SDK. Programmatic integration for Python applications.
- n8n Node. Official node for adding compliance gates to n8n workflows.
- MCP Server. Model Context Protocol integration for compatible systems.
Trace Lifecycle
Follow a trace from capture to final state.
Review Tasks
Understand the human review process.
Channels
See where notifications and decisions happen.