Cloud/AI Governance/Mission Control
Mission Control
The governance command center: every request, every millisecond, every redacted leak and every truncated payload across your AI fleet, on one screen, with an AI briefing that tells you where to look first.
Mission Control is the first screen of AI Governance, and it answers the question every AI operator asks: is my AI operation healthy right now? Every request your agents make through Vinkius lands here as a metric, a chart or a live row. You see the whole fleet in one glance, and every card points to the exact report for anything that needs a closer look. This is the difference between using AI and running AI in production.
Mission Control
Know exactly what needs your attention before you even look at the charts. One briefing. The full picture. The next move.
AI Agents Activities
Interactions per hourRequest Volume & Latency
Hourly dual-axis overviewTop MCP Servers
Volume by server · color = latency healthFirewall Enforcement
Hourly DLP + FinOps actionsCost vs Savings
Hourly token cost attributionFailure Breakdown
Hourly requests · Agent / Upstream / Vinkius failuresTool Health Matrix
Latency × Failure rate · bubble size = call volumeLive Activity
Latest tool executions across all MCP servers| MCP Server | Tool | Action | Outcome | Total |
|---|---|---|---|---|
| GitHub | list_issues | Query | Enforced200 | 96ms |
| Jira | create_task | Mutation | OK200 | 74ms |
| Notion | search_docs | Query | Enforced200 | 63ms |
| HubSpot | get_user | Query | Agent Err401 | 41ms |
| Slack | send_message | Destructive | Upstream Err503 | 132ms |
| Supabase | fetch_resource | Query | Vinkius Err500 | 88ms |
The mockup above shows the console's own Sample Data, the same simulated dataset the free tier explores with. On a paid plan the exact same screen renders your operation's live numbers, per organization. Nothing in your dashboard is a fixed number: every KPI, chart and row is your own traffic, refreshed on demand.
The screen from top to bottom
- Period control. Today, 3 days, 7 days, 14 days or 30 days. Today and 3 days break down hour by hour; longer periods break down day by day. The refresh button revalidates everything on demand.
- AI Briefing. An AI that reads your own numbers and writes the executive summary: what needs attention before you even look at a single chart. The console describes it in one line: "Know exactly what needs your attention before you even look at the charts. One briefing. The full picture. The next move." On paid plans it generates a live briefing of your fleet; on the Free plan you see the description with an Upgrade to unlock button.
- Six KPIs, each with a health dot and a sparkline. Requests (total in the period), Avg Latency (split into the upstream API time and the Vinkius overhead, so you can see what the proxy adds: it is milliseconds), Vinkius Reliability (with failures split into Agent, Upstream and Vinkius), Tokens processed, DLP Protected (sensitive data redacted) and Cost Saved (FinOps truncations applied). The dots are not decoration: latency is green under 200ms, amber under 1s, red after that; reliability is green at 99% or higher, amber from 95%. Problems announce themselves.
- Seven charts, each a door. AI Agents Activities (heatmap of interactions per hour or day), Request Volume & Latency (dual axis), Top MCP Servers (volume by Connector, color is latency health), Firewall Enforcement (DLP redactions plus FinOps truncations over time), Cost vs Savings (token cost attribution), Failure Breakdown (Agent / Upstream / Vinkius failures) and Tool Health Matrix (latency versus failure rate, bubble size is call volume). Every chart has Details to open its dedicated report and a maximize button to expand it in place.
Live Activity: the fleet, row by row
At the bottom, under a pulsing green dot, the Live Activity feed streams the latest tool executions across all MCP servers: which Connector, which tool, what kind of action (Query, Mutation or Destructive), the latency and the outcome.
The outcome is where Vinkius is different: every row names who is responsible.
| Outcome | Meaning |
|---|---|
| OK | The request succeeded as-is. |
| Enforced | It succeeded and Vinkius policy applied in flight: DLP redactions, FinOps truncation of an oversized payload, or both. Protection you can audit, request by request. |
| Agent Err (4xx) | The calling agent sent a bad request. The console shows the exact reason, for example "Unauthorized: token lacks required scope for issues:read". |
| Upstream Err (5xx) | the MCP provider failed, for example a rate limit: "Service Unavailable: upstream rate limit exceeded (retry after 60s)". |
| Vinkius Err | A failure originated inside the Vinkius proxy itself, and the feed says so explicitly. When the proxy blocks an attack it also signs it: "SSRF_BLOCKED: host 169.254.169.254 resolves to a private IP range". |
No more "the AI integration is failing" and a day of guessing. One look, one row, one name.
Free plan: explore the whole command center
On the Free plan Mission Control runs on Sample Data (the amber badge next to the title explains it, and hovering it says: "This dashboard is populated with simulated data so you can explore every chart and metric"). You learn the entire control center before paying anything. A paid plan switches the same screen to your live fleet, per organization, and unlocks the real AI Briefing.
From the overview to the answer
Mission Control is the summary; the eight Reports screens are the deep dives, and every chart's Details link takes you straight to the right one: Agent Activity, Server Traffic, Access Tokens, AI Spend, Tool Reliability, Security Posture and Request Failures. The four Settings screens (Connector Policy, DLP Protection, FinOps Guard, Circuit Breaker) are where you decide what the fleet is allowed to do in the first place. Each of them gets its own page in the sidebar, in the console's order.