> ## Documentation Index > Fetch the complete documentation index at: https://app.keystroke.ai/docs/llms.txt > Use this file to discover all available pages before exploring further. # Overview > Inspect agent and workflow runs across your projects. Run history is where you debug and audit what agents and workflows did. Use it to see whether a run finished, what started it, which tools or models it used, what it cost, and where it failed. Open the **History** tab in the web app (in main sidebar) to see runs across the projects you can access in the active organization. ## What appears in History History includes two run types: | Type | What it represents | | ------------ | --------------------------------------------------------------------------------------------------------------------- | | **Agent** | A prompt session with an agent, whether started from the API, CLI, a workflow, another agent, a trigger, or a gateway | | **Workflow** | A workflow execution, whether started directly, by a trigger, or from another runtime path | The table shows each run's name, type, start time, actor, status, trigger, duration, and project. Additional columns are available for run ID, queue time, finish time, models used, other services used, total cost, message count, step count, and raw trigger value. You can search runs, filter by fields like status, type, project, trigger, actor, model, service, cost, messages, and steps, or use the saved views for failed and running runs. ## Statuses Runs can have these statuses: | Status | What it means | | ------------------- | ------------------------------------------------ | | **Queued** | The run has been created and is waiting to start | | **Running** | The run is actively executing | | **Sleeping** | The run is paused until the runtime wakes it | | **Waiting on hook** | The run is waiting for an external hook response | | **Success** | The run completed successfully | | **Failed** | The run ended with an error | | **Canceled** | The run was canceled before it completed | Active runs update automatically while they are queued, running, sleeping, or waiting on a hook. ## Inspect a run Select a row in **History** to open the detail panel. The panel has three tabs: | Tab | What it shows | | ------------ | ------------------------------------------------------------------------------------------------------------------ | | **Overview** | Summary, conversation or payloads, usage, and run cost details | | **Metadata** | Trigger time, queue time, start and finish times, status, run ID, actor, project, models, services, and total cost | | **Trace** | Raw trace spans and logs in JSON or YAML | Agent runs show the recorded conversation in the Overview tab. Workflow runs show input and output payloads, or an error payload when the run failed. The Trace tab is the lowest-level view. Spans nest to mirror what actually ran: a trigger run wraps the workflow run, which wraps each action, agent, and model step, which in turn wrap nested tool calls. Reading the spans shows what ran, in what order, and where time and errors went. ## Use the CLI Use `keystroke history` when you want organization-level cloud history from the terminal: ```bash theme={null} keystroke history list --kind workflow --status failed keystroke history get ``` `keystroke history list` uses the active cloud organization and can filter by project ID, status, and kind. Owners and admins can pass `--admin` to list all organization resources. For resource-specific inspection, use the agent and workflow run commands: ```bash theme={null} keystroke agents sessions list support keystroke workflows runs list signup-pipeline ``` Those commands follow the resolved CLI target, so they can inspect local runs while developing or cloud runs after deploy. See [targets: local vs cloud](/cli#targets-local-vs-cloud) and the [CLI reference](/cli#invoke-and-inspect). ## Next steps Inspect agent sessions, conversations, tool calls, and traces. Inspect workflow input, output, errors, steps, and traces. Start agent sessions and inspect their results. Start workflow runs and review their output.