Getting startedSet up integrationsCreate your first initiativeInvite your teamPlan today's workShare your first update
OverviewQuickstartWorkflowsLocal AgentsMonitoringUse in terminalTroubleshooting
DocsAPI Reference

Main

  • Home
  • About
  • Pricing
  • Vault
  • Changelog
  • Docs

Features

  • Roadmaps
  • Planning
  • Standups
  • Status updates
  • Insights
  • AI assistant / MCP
  • Integrations

Solutions

  • Startups
  • Dev shops / agencies
  • Software teams
  • Internal IT & platform teams

Alternatives

  • vs Jira
  • vs Linear
  • vs Asana
  • vs Monday.com
  • vs ClickUp
  • vs Notion

Company

  • Blog
  • Security
  • Log in
  • Sign up
  • Terms of Use
  • Privacy Policy

Resources

  • Docs
  • Community
  • API reference
  • CLI
  • Desktop app
  • SDK

© 2026 One Horizon. All rights reserved

FacebookInstagramThreadsXRedditTikTokYouTubeMedium


Monitor workflows and agents

Monitor execution from Workflows, Work queue, the linked work item, or an agent details page. Each view answers a different question without changing the underlying run.

Choose the right view

ViewWhat to inspect
Workflows overviewDefinitions, online agents, recent sessions, and workflow runs.
Workflow Runs tabStep attempts, routes, review gates, stalls, branches, and terminal outcomes for one definition.
Work queueUp to ten queued, active, awaiting-input, and recently finished sessions without leaving the current page. The badge counts queued and active sessions.
Session detailsClaim owner, activity timeline, provider metadata, token usage, worktree, errors, and available actions.
Agent detailsOnline, stale, or offline health; settings; sessions targeted to that agent; and local start or restart controls.
Linked work itemThe specification, workflow documents, comments, branch or PR/MR, and final delivery context.

Open a session when you need to distinguish queued work from an active claim. Inspect the status, latest activity, and agent heartbeat before choosing a recovery action.

The session detail view brings claim state, provider metadata, token totals, and actions together.

Session states

StateOperator meaning
queuedWaiting for an eligible agent to claim it.
pendingAccepted but not ready to execute.
activeAn agent holds the claim and is executing.
awaiting_inputThe agent paused for a reply. Choose Resume and provide the requested input.
completeThe agent reported success. Review the result before merging or closing the work.
errorExecution failed. Inspect the error and activity before retrying.
staleHeartbeats or the claim expired. The previous agent may reclaim it, or you can create a replacement run.
cancelledA user or system cancelled the session.

Workflow runs separately use not_started, running, waiting_for_review, stalled, completed, cancelled, and failed. A session can finish while its workflow waits at the next step or review gate.

Actions

ActionWhen it appearsResult
ResumeAn agent session is awaiting_inputQueues one reply for the owning online agent and refreshes the active claim lease.
Resume on a workflow runA stalled run exhausted the current step's pass limitSkips the exhausted step attempt and follows that step's configured route.
Try againA session is stale, errored, cancelled, queued, or pending; completed sessions can also be retriedCreates a replacement session with the same work, prompt, agent profile, and target. Failed or unstarted source sessions are archived after replacement.
ArchiveA session is complete, error, cancelled, or staleHides it from queue and session lists without changing the work item or its activity history.
CancelA session or workflow run is still activeStops the current execution path and records a cancelled outcome.
Open in CodexA Codex local session has a recorded thread IDOpens the recorded Codex thread.
Open in Claude CodeA Claude Code local run is complete or errored and has a session IDResumes that session in the task workspace folder.

Pass-limit recovery changes the route the workflow will take next.

Resume after a pass-limit stall skips the exhausted step and follows its configured route. Inspect the stalled reason, current attempt, and route destination before confirming.

For an awaiting-input session, the reply can contain up to 20,000 characters. Only one reply can be pending, and the owning agent must be online. The agent consumes the input on its next poll without creating a new claim.

Review gates

Human review steps appear in Work Inbox and the workflow run. Approve continues forward, feedback returns to the configured earlier step, reject follows the explicit reject route, and reassignment changes the reviewer without advancing the run.

Agent review steps make the same approve, feedback, or reject decision. A missing verdict stalls the workflow instead of assuming approval. A reported verdict can route the workflow, but activity text alone is not a verdict and the agent session still needs a terminal state to be complete.

Activities and metadata

Activities explain meaningful progress. Common entries include progress, plan_updated, external_url_updated, awaiting_input, user_resume_input, completed, failed, and policy_decision. Heartbeats indicate health and should not be read as progress.

Session metadata can include provider thread and process IDs, model and provider, working directory, worktree path, provider status, external URL, and token counts. Session details show total, input, and output token counts after a terminal state when the runtime reported them.

Agent health

Local agents heartbeat while their process is running. A missed heartbeat moves the agent through stale and offline health, expires active claims, and can leave sessions stale. Check Desktop ownership, Auto start, agent logs, one agent status, and local resource pressure before replacing the run.

Recently finished sessions remain in Work queue for about eight hours before they age out. Archive removes a finished session immediately; archived sessions cannot be claimed or restored from the dashboard.


PreviousLocal AgentsNextUse in terminal

Workflows

Create, publish, launch, and operate workflows with agent steps and human review gates.

Reference

Reference runtime responsibilities, workflow context, artifact documents, environment variables, and API operation families.

Run & monitor

Operating a custom runtime in daily use: process model, compatibility levels, dashboard monitoring, and logs.

Troubleshooting

Fix setup, publishing, launch, connectivity, execution, review, Git, and handoff failures.

  • Choose the right view
  • Session states
  • Actions
  • Review gates
  • Activities and metadata
  • Agent health
  • Back to top