Getting startedCreate your first initiativeConnect version controlInvite your teamPlan today's work
OverviewQuickstartWorkflowsCapabilitiesLocal AgentsMonitoringUse in terminalTroubleshooting
DocsAPI Reference

Main

  • Home
  • About
  • Pricing
  • Changelog
  • Docs

Features

  • Roadmap
  • Boards
  • Triage
  • Workflows & agents
  • Progress
  • Insights
  • CLI
  • 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
  • Desktop app
  • SDK
  • Vault
  • QR code generator

© 2026 One Horizon. All rights reserved

FacebookInstagramThreadsXTikTokYouTubeMedium


Run your first workflow

Desktop is the fastest path from a work item to a running agent. Have a Git repository ready and sign in to Claude Code, Cursor, or OpenCode before setup if you plan to use one of those runtimes. Codex is bundled. For OpenCode, follow the OpenCode setup guidance, then confirm opencode is available on PATH and the provider session is authenticated.

  1. 1

    Open Desktop

    Install and open One Horizon Desktop, then sign in and choose the workspace that owns the work.

  2. 2

    Choose the automation level

    Open Workflows and choose the guided setup that matches the work. Start with a simple build workflow if you want one agent to implement a task, or include review steps when approval must happen before the run continues.

  3. 3

    Select a runtime and repository

    Choose Codex, Claude Code, Cursor, or OpenCode, then select the repository folder the agent may use. Claude Code, Cursor, and OpenCode require authenticated local CLIs. OpenCode supports code work, so choose a Builder workflow. The selected repository belongs to this agent on this machine.

  4. 4

    Complete setup

    Review the model, agent access, code isolation, workflow file, and pull request or merge request settings. Finish setup and wait until the agent shows as online. From the agent's Capabilities card, add Agent Skills or MCP servers. Claude Code, Cursor, and OpenCode agents can also add plugins; Codex agents cannot. Agent capabilities covers catalog and connection requirements.

  5. 5

    Launch with Send to…

    Open an initiative, bug, or todo that contains a usable specification. Select Send to…, choose the published workflow, and confirm the run.

    Inspect workflow names at the top of the menu, direct agent targets below them, and app launch targets for a prefilled prompt. Send to Agents explains Send to workflow, Send to agent, and Send to app.

  6. 6

    Verify the run

    Open Workflows or Work queue and confirm the first step is active or queued for the selected agent. The linked work item keeps the run, activities, documents, branch, and review result.

The run may pause for human review, agent input, or a configured route. Monitoring explains the available actions for each state.

Review generated changes, tests, documents, and any pull request or merge request before merging. A completed agent session is delivery evidence, not automatic approval.


PreviousOverviewNextWorkflows

Overview

Choose how to run built-in agents and workflows from Desktop, the dashboard, or the CLI.

Send to Agents

Send work into a workflow, assign it directly to an agent, or open a coding app with a prefilled prompt.

Capabilities

Add skills, MCP servers, and plugins, then control which agents and workflow steps can use them.

Local Agents

Create and run local agents from Desktop, with optional terminal control.