Runs & Observability

AKOS captures detailed telemetry for every execution. The Runs, Traces, and Cost screens give you full visibility into what happened, how long it took, and what it cost.

Runs screen

The Runs screen lists every workflow execution in the workspace, most recent first.

Run list

Each row shows:

Run ID and task description.
Status badge: running, completed, failed, paused (waiting for human approval), or cancelled.
Duration, cost in USD, and token count.
The trigger that started this run (cron name, webhook path, manual, etc.).

Use the search box to filter runs by task text or run ID. An overflow actions menu provides:

Compare runs — select two runs to diff their outputs side by side.
Export CSV / NDJSON — download the visible run list (requires runs:export permission).
Export signed batch — download a cryptographically signed export for compliance.
Print — browser print of the visible list.

If a run was started by a specific trigger or agent, a filter chip appears above the list. Click Clear to remove the filter.

Run detail panel

Click any run to open the detail panel:

Header

Run ID and task name.
Status icon and a copy-share-link button.
A View as artifact toggle that renders the run's output in the artifact viewer.
Run controls (stop, retry — where available).

Metrics

Duration, cost (USD), token counts (prompt + completion), and the trigger name.

Assurance

Governance checks: egress rules evaluated, firewall decisions, sandbox mode.

Agents

Each agent that participated, with its provider, model, and per-agent status.

Tools used

Skill and tool calls made during the run.

Evals / Benchmarks

Any automated evaluation scores attached to this run.

Artifacts

Documents, code files, images, or other outputs the run produced. Click an artifact to preview it in the Assets screen.

Tool calls

A detailed log of every tool invocation, with input and output.

Traces screen

The Traces screen shows OpenTelemetry-compatible distributed traces for runs. Each trace is a tree of spans — one span per agent call, tool invocation, or sub-flow.

Use the Traces screen when you need to diagnose latency or unexpected behaviour at the span level. Click any span to see its attributes, duration, status, and any error messages.

Observability screen

The Observability screen provides aggregate monitoring views:

Health — system-level health checks for the sidecar, storage, and any configured services.
Metrics — charts for run volume, error rates, latency, and cost trends.
Exporters — configure where AKOS forwards its telemetry (see the Connections screen for setup).

Cost screen

The Cost screen breaks down AI spending by agent, provider, workflow, and time period.

Summary — total spend this period, budget consumed (as a percentage), and the top cost drivers.
By agent — a ranked list of agents sorted by spend.
By provider — spend grouped by LLM provider.
Timeline — a day-by-day or hour-by-hour bar chart.

Budget limits and alerts are configured in the Workspace settings (see Workspaces).

In the flow editor, click Trace in the canvas toolbar to overlay the last run's telemetry directly on the canvas. Each node shows its status icon (passed, failed, skipped), duration, and cost. This is the fastest way to pinpoint which step in a flow is slow or failing.