Audit and History

Business Summary: Audit & History records every significant platform event — agent actions, human interactions with agents, MCP reads and writes, agent runs, and approvals — so compliance teams can trace causality and query the same data from Perception at any time.

Open Audit & History from the sidebar (Governance section). Every event links to the Lifecycle Graph via View Lineage.

On this page

• Start here — task guides
• What gets recorded
• Decision traces
• Integration audit
• Workstream Activity tab
• Example scenario
• FAQs

How this fits in Nimbus

flowchart LR
  Run[Agent run]
  Exec[Execution events]
  Gov[Governance events]
  Export[Export events]
  Conn[Connector events]
  Activity[Workstream Activity]
  Audit[Audit and History]
  LG[Lifecycle Graph]

  Run --> Exec
  Run --> Gov
  Run --> Export
  Run --> Conn
  Exec --> Activity
  Gov --> Activity
  Export --> Activity
  Conn --> Activity
  Activity --> Audit
  Audit -->|View Lineage| LG

Recording happens at every layer of governed work. You do not configure what gets logged for standard governed runs — the platform writes events as work progresses.

Start here {#start-here}

What gets recorded {#what-gets-recorded}

Nimbus treats accountability as a product feature, not an afterthought. When agents run inside a workstream, the platform records enough detail to reconstruct what happened, who was involved, and what was released — without asking you to keep separate notes.

Execution, governance, export, and connector events aggregate into Workstream Activity and org-wide Audit & History as shown in the diagram above.

Execution events

These record what ran and what it produced. When you confirm a run plan and the agent executes, Nimbus timestamps the start and completion, links the objective you approved, and records artifacts that land on the Outputs tab. NTU spend is attributed to the run so finance can reconcile usage later.

Typical execution events include run started and completed, plan confirmed, artifacts landed, and NTU spend tied to a run.

Governance events

These record who decided and what gates applied. When the agent pauses at a human-in-the-loop (HITL) checkpoint, Nimbus records the pause, your sign-off action, and the timestamp. Dual approver steps and spend or connector gates appear here too. See Checkpoints and approvals.

Export and commit events

These record what left Nimbus or what changed downstream — exporting an approved report, committing a decision to a CRM, or syncing documents into the Company Wiki.

Connector and integration events

These record data in and out of connected systems. Intelligence connector reads show what informed a recommendation. Action connector writes appear only after governance gates clear, with the approving actor attached.

Configuration changes

These record who changed workspace setup — connector connections, workstream creation, and governance preset updates. Some changes are visible today in workstream Activity and connector status on Connectors in the sidebar.

What is not recorded in user-facing history

Nimbus does not surface raw model prompts or full chat transcripts in standard audit views. The decision trace summarizes evidence, agents, and approvals in plain language.

Decision traces {#decision-traces}

A decision trace is run-level evidence written in plain language. It answers the questions executives and auditors ask most often: What did the AI actually do? What data did it rely on? Who signed off before anything went external?

flowchart TB
  Plan[Confirm and run]
  Run[Agent execution]
  HITL[HITL checkpoints]
  Trace[Decision trace]
  Artifact[Approved artifact]

  Plan --> Run
  Run --> HITL
  HITL --> Run
  Run --> Trace
  Run --> Artifact
  Trace --> Activity[Activity tab]
  Artifact --> Outputs[Outputs tab]

When a run completes HITL and outputs are committed, Nimbus records a durable entry in the operating ledger — analogous to a Git commit. The trace captures the evidence chain at commit time.

What a decision trace contains

The trace is immutable once recorded. If you need a different outcome, start a new run with a refined objective rather than editing past evidence.

Where to find decision traces

• Workstream Activity tab — expand run completed events to see trace summaries and linked artifacts.
• Workstream chat — HITL interrupt cards and run completion messages reference the same run ID.
• Overview — the Recent decisions table highlights committed outcomes; pending items appear under Pending approvals.
• Full workstream studio — Open studio from the workstream header exposes the richest trace view.

Org-wide search across all traces will eventually live on Audit & History. Today, open the workstream where the run executed.

Decision traces vs chat history

Perception Chat history (workspace-wide) is separate from workstream chat — sessions persist for your Perception queries; workstream chat threads stay scoped to you inside each workstream.

For compliance conversations, cite the trace and approved artifacts — not informal chat fragments or another user's private threads.

Integration audit {#integration-audit}

Connectors link Nimbus to CRMs, data warehouses, document stores, and messaging tools. Because agents can read sensitive data and, with approval, change downstream records, every integration action leaves an audit footprint.

flowchart LR
  Systems[External systems]
  Intel[Intelligence connector]
  Agent[Agent run]
  Action[Action connector]
  Activity[Workstream Activity]
  Trace[Decision trace]

  Systems --> Intel --> Agent
  Agent --> Action --> Systems
  Intel --> Activity
  Action --> Activity
  Agent --> Trace

Intelligence connectors read data for analysis without changing the external system. Action connectors write or trigger operations only after governance gates clear — Critical HITL for external writes, spend thresholds, or dual approvers when configured. The audit record includes the approving actor, not just the agent.

Connections belong to your organization — one active link per connector ID, shared across workstreams. Audit entries still record which workstream and run used the connection so you can trace scope.

Workstream Activity tab {#workstream-activity-tab}

Every committed workstream keeps its own audit trail. While the org-wide Audit & History page is still coming online, the Activity tab is usually the best place to answer "what happened on this workstream?"

flowchart LR
  Chat[Workstream chat]
  Run[Governed run]
  Outputs[Outputs tab]
  Activity[Activity tab]
  Trace[Decision trace]

  Chat --> Run
  Run --> Outputs
  Run --> Activity
  Run --> Trace
  Activity --> Trace

Opening Activity

1. Go to Workstreams in the sidebar.
2. Open a workstream card.
3. Select the Activity tab alongside workstream chat, Outputs, and Milestones.

Activity is a reverse-chronological log — newest events at the top.

Activity vs Outputs vs Milestones

Outputs shows what was produced — artifacts you can open and export. Milestones shows progress toward the goal and approval gates along the way. Activity shows how it happened — chronological evidence. Approved artifacts on Outputs should have matching Artifact landed and Output approved events on Activity.

For the richest timeline — including decision traces — open Open studio from the workstream header. Team members with access to the workstream see the same Activity tab.

Example scenario

Your compliance team asks whether a quarterly board pack used live ERP data or a stale spreadsheet. You open the workstream Activity tab and find connector read events around the run date. You open the linked decision trace and confirm the evidence section lists the warehouse connector. You also see a Plan confirmed event with the objective your FP&A lead approved. That chain — plan, read, trace — answers the audit question without exporting raw chat.

FAQs {#faqs}

Q: Can I query audit data from Perception? A: Yes — ask about approvals, MCP writes, or run outcomes in natural language. The Perception orb queries the same operating memory as Audit & History. See Using the Perception console.

Q: How does Audit connect to the lifecycle graph? A: Click View Lineage on any event to jump to related graph nodes.

Q: Where do HITL approvals appear? A: As governance events with the approving actor — also visible in Teams HITL counts.

Q: Do I need to turn recording on? A: No. Recording is automatic for governed runs inside workstreams.

Q: Is Activity the same as a decision trace? A: Activity is the timeline you browse. A decision trace is the structured evidence behind a specific run — agents, evidence, approvals.

Q: Does disconnecting a connector erase audit history? A: No. Past run and Activity records remain. New runs cannot use the connector until you reconnect.

Q: Are informal chat messages part of the audit record? A: Workstream chat is private per user and not exported to org-wide audit views. The decision trace, Activity timeline, and approved artifacts capture governed execution for auditors.