The Problem With AI Output
When an AI agent does research, writes a summary, or analyses a dataset, the result typically lives in one of two places: a chat window that disappears when you close it, or a file you saved somewhere and will probably never find again.
Neither is acceptable for a production AI workforce. You need a permanent, structured, searchable record of what every agent produced, why, and when. That is what Nexus Reports provide.
What Is a Nexus Report?
A Nexus Report is a structured Markdown document automatically generated at the end of every agent cascade run. It captures:
- The triggering event — what caused this cascade to run (schedule, event, manual trigger)
- The cascade steps — each step, its inputs, and its outputs in sequence
- The agent output — the full content produced by any agent step, including model attribution
- The metadata — timestamp, duration, report ID, cascade name, and the agent persona that ran it
Reports are filed automatically to your reports history. You do not create them — they are a by-product of the cascade running. The system generates them whether the run succeeded, partially succeeded, or failed.
The Reports Interface
Reports are browsable from the Mission Control reports panel:
Per-persona filtering — Click any agent persona to see only reports produced by that agent. Your Librarian's research output is separate from your Cipher's security assessments, separate from your Medic's system diagnostics.
Search — Full-text search across report content, not just titles. If you remember that an agent mentioned a specific company or technical term, you can find the report that contained it.
Date range filtering — Narrow to a specific day, week, or custom range.
Step-level drill-down — Each report expands to show individual cascade steps with their inputs and outputs. You are not just reading the final summary — you can see every intermediate result the cascade produced.
Model attribution — Every piece of AI-generated content in a report is tagged with the provider and model that produced it. You always know which model wrote which section.
Deep-Linked Notifications
Reports connect to the notification system. When a cascade completes, the notification that appears in your activity sidebar links directly to the specific report that run produced. Click the notification, land on the report — no searching required.
This makes the notification-to-audit flow instantaneous. Security event? Click the Discord notification. You are reading the full Nexus Report for that event within seconds.
Nexus News
The News module extends the reports concept into real-time intelligence feeds.
A News Feed is a configured set of sources — RSS feeds, URLs, search queries — that the platform monitors on a schedule. When new content appears, a cascade processes it: summarising, tagging, extracting key claims, and flagging items that match your configured interest criteria.
The output lands in your News dashboard as a curated briefing — not a raw RSS reader, but an AI-filtered view of what matters from the sources you care about.
Configuring a News Feed
feed:
name: "AI Governance Weekly"
sources:
- type: rss
url: "https://venturebeat.com/category/ai/feed/"
- type: rss
url: "https://techcrunch.com/tag/artificial-intelligence/feed/"
- type: search
query: "AI regulation EU 2026"
schedule: "every Monday at 07:00"
filter:
keywords: ["regulation", "governance", "liability", "safety"]
min_relevance: 0.7
agent: librarian
output: nexus_report
The Librarian agent processes each new item, scores it against your keywords and relevance threshold, and files a Nexus Report with the curated briefing. Monday morning, your report is waiting.
Reports as Audit Trail
For governed AI systems, reports are not just convenient — they are essential.
Every ban decision, every content moderation call, every automated action has a Nexus Report behind it. If a decision is questioned, you pull the report: here is the data the agent saw, here is the assessment it produced, here is the model that produced it, here is the timestamp.
This is what distinguishes the Sovereignty Protocol from a standard automation platform. Automation platforms tell you what happened. The Sovereignty Protocol tells you what happened, who decided it, why, and with what model — permanently, without manual effort.
Integration With Cascades
Reports are generated automatically by the cascade runtime. You do not need to add a "generate report" step to your cascades — the system produces a report for every run by default.
You can also write a cascade that explicitly creates or appends to a report:
- name: file_report
type: db_write
collection: nexus_reports
data:
title: "Weekly Competitive Intelligence Brief"
content: "$synthesis.content$"
persona: "librarian"
cascade: "$meta.cascade_name$"
generated_at: "$meta.timestamp$"
This lets you customise the report format, combine outputs from multiple cascade steps into a single document, and control exactly what gets filed.
Your AI workforce produces output. Nexus Reports make sure you can always find it.