Runs
Where: Sidebar › Agents › Runs. The badge shows how many runs are executing right now. A live table of every agent execution — auto-refreshing every 5 seconds (a green pulsing indicator shows this). Columns include Agent, Status, Level, Department, Project, Trigger, Started, Duration, and Summary (more columns like Error, Self-Healed, and Run ID can be toggled on).- Stat cards: Total / Success / Failed / Running / Agents.
- Filter presets: Success, Failed, Running, Pending, Error — plus the standard table toolbar (search, filters, grouping, time range, CSV export; see Data Views).
- Run statuses you will encounter: running, success, failed, budget exceeded, waiting for user, paused, approval timeout, rejected, and error variants.
Run detail
Click a run to open Run Detail:- Header: run ID, status, agent, trigger and start time, plus Duration / Cost / Tokens (cost is marked estimated unless confirmed by the provider).
- Input, Output, and Error sections with expanders.
- Tabs: Timeline (run started → each tool call with status and timing → run finished), Raw Logs, Telegram (for Telegram-triggered runs, with sensitive fields masked), and Cost (total, credit charge with its credit class, token breakdown, per-step list).
- Replay re-executes the run; the new run is linked to the original.
Agent Decision Log
Where: Sidebar › Agents › Agent Decision Log. An audit trail of agent decisions: every tool call, whether the agent acted autonomously, its reasoning, escalations, and self-heal events. Columns: When, Agent, Project, Type, Auto, Self-heal, Action, Reason, Escalated, and more. This log is also what powers “receipts” — an agent that claims work without recorded tool calls is flagged as an unverified claim. Auto-refreshes every 5 seconds.Agent Inbox
Where: Sidebar › Agents › Agent Inbox. The badge counts unread agent messages. The agent-to-agent message log — what your agents send each other (handoffs, delegated tasks, reports up the chain). Columns: Created, Subject, From, To, Type, Body, Tags. It is a read-only table view — your own actionable inbox is the company Inbox. This is where delegation becomes visible: when an agent hands work to another agent, the message sits here unread until the recipient’s next run, at which point it is picked up and injected as a pending task (“address these first”) and marked read. Reports flow the other way — up the reports-to chain until the CEO agent digests them into your Inbox. Runs that one agent delegates to another appear as children of the parent run in the Runs view, so the whole chain stays traceable end to end (see Agents › How delegation works).Analytics
Where: Sidebar › Agents › Analytics. A dashboard of cost, performance, and reliability. Pick a range (7d / 14d / 30d / 60d / 90d) and refresh on demand.- KPI cards: Total Runs, Total Cost, Avg Duration (with p50/p95), Success Rate.
- Cost Trend (30d) bar chart and Token Usage (input vs output split).
- Top Expensive Agents — cost, runs, avg per run, and share of spend.
- Failed Runs — the last 50 failures with agent, status, error, cost, and time.
- Budget banners warn when company spend passes 80% (Budget warning) and 95% (Budget critical, with a projection of when the budget will be exceeded).
Cost
Where: Sidebar › Agents › Cost. The Agent Cost Dashboard: pick a period (This Month / Last Month / Last 7 Days) and see:- The Monthly Budget bar (spent / budget / remaining, colored green → orange → red).
- Cost / Day (30d) chart and summary stats (Period Spend, Active Agents, Budget Utilization).
- Cost by Agent — sortable table with total cost, runs, Saved (the ROI you configured per agent), top trigger chips, share of spend, and a Drill link that opens a per-agent breakdown (cost by trigger, most expensive runs).
- Top 10 Expensive Runs.
- Each figure carries a cost-source badge: Real (provider-confirmed) vs estimated.