The Problem With Invisible AI Work
When AI agents do work in the background — research, enrichment, code review, data cleanup — the default experience is a black box. You run the job. You wait. You get output. You have no idea what happened in between.
This is a governance problem as much as a usability problem. If you cannot see what your agent workforce is doing, you cannot verify it is doing the right things. You cannot improve the work. You cannot catch problems early. You cannot give meaningful credit to the agents that are performing well.
Nexus Mission Control exists to make the AI workforce visible.
The Operations Tab
The heart of Mission Control is the Operations tab — a live feed of cascade runs, with full drill-down into every step.
For each cascade run you can see:
- Which cascade ran — name, persona, trigger type (schedule, event, manual)
- When it ran — timestamp, duration, and whether it is still in progress
- What each step did — the full output of every step in the cascade, expandable inline
- Which model was used — the provider and model ID for each agent step
- Errors and skipped steps — any step that failed, why it failed, and whether it was marked optional
This is not just a log. It is a structured record of every decision the cascade made, in sequence, with context. If a cascade produced an unexpected output, you can trace it back to exactly which step and which model response caused it.
The Tasks Tab
The Tasks tab is a JIRA-style Kanban board for agent-generated work items.
Lanes:
- Intake — new tasks waiting for assignment or agent pickup
- In Progress — tasks currently being worked on by an agent
- Done — Evaluation — tasks completed but awaiting human review
- Archived — reviewed and closed tasks
Tasks support:
- Tags and keywords — for filtering and categorisation
- Priority — urgent, high, normal, low
- Due dates — with visual overdue indicators
- Sub-tasks — recursive nesting for complex items
- Agent assignment — which agent persona owns this task
- Schedule pills — tasks generated by loop cascades show their next scheduled run time
Agents can create tasks via MCP tool call, move them between columns autonomously, and mark them complete — all logged and auditable. You can also create, edit, and move tasks manually.
The Workforce Tab
The Workforce tab shows the named agent personas — Librarian, Linter, Medic, Cipher, and others — with their current XP, level, and recent work history.
For each agent you can see:
- XP total and current level — earned from completed tasks, runs, and approved outputs
- Recent activity — a timeline of the last N tasks this agent completed
- Work output previews — click any activity entry to see the full output
- Award bonus XP — manually award XP for exceptional work visible directly in the dashboard
The workforce view makes the agent team feel like a real team. You know which agent is most active, which ones are producing high-quality output, and which ones have been underutilised.
Nexus Reports Integration
Every cascade run, every enrichment, every research task automatically generates a Nexus Report — a structured Markdown document filed to your reports history.
Reports appear in Mission Control's reports panel with:
- Per-persona tabs — filter reports by which agent produced them
- Search and filter — find reports by date, persona, cascade name, or content keywords
- Full content view — rendered Markdown with collapsible step outputs
- Model attribution — which model produced which content in the report
Deep-linked notifications connect to reports directly — click a "cascade completed" notification and land on the specific report that run produced.
Real-Time Activity
Mission Control includes a real-time activity sidebar that shows:
- Running agents — which agents are currently active with a live status indicator
- Recent completions — the last N agent tasks to finish, with a quick-look preview
- Error alerts — any cascade step that failed in the last 24 hours
- XP events — a live feed of XP awards as they happen
The sidebar is visible from any page in the dashboard, so you always have a peripheral view of what your workforce is doing without needing to navigate to Mission Control explicitly.
Why One Dashboard Matters
The alternative to Mission Control is context-switching between logs, output files, notifications, and separate monitoring tools. That works for small deployments. It does not scale.
As your agent workforce grows — more cascades, more personas, more background tasks — the cognitive overhead of tracking everything separately becomes unmanageable. Mission Control centralises all of it: operations, tasks, workforce status, reports, and notifications in one place, with the drill-down depth to investigate anything unusual.
This is what it looks like to run an AI workforce instead of just running AI.