Kanban Mode
Delegate coding tasks with visible review gates
Turn prompts into task cards, run a shared agent or Swarm, and keep final responses, transcripts, and tool timelines attached for review.
Local-first AI coding agent orchestration
Dorchestrator
Dorchestrator brings Claude Code, OpenAI Codex, and built-in coding agents into one desktop workspace. Plan and review tasks on a Kanban board, coordinate multi-agent swarms, reuse terminal layouts and Skills, and inspect readable tool activity with workflow state stored in your selected project folder.
- 4
- workspace modes
- 10
- visual themes
- 1
- control layer
Product
Four workspaces for orchestrating AI coding agents
Use Kanban to delegate and review tasks, Swarm to coordinate connected agents, Mux to manage repeatable terminal layouts, and Skill to install reusable SKILL.md capabilities. Every mode works against the same selected local workspace.
Swarm Mode
Coordinate multi-agent coding workflows
Build graph-based agent teams with reusable roles, connected messaging, shared project context, and provider-specific terminal choices.
Mux Mode
Run repeatable multi-terminal agent sessions
Save template-driven terminal layouts for recurring workflows, then switch quickly between shell, Claude Code, Codex, and built-in agent contexts.
Skill Mode
Manage reusable coding-agent Skills
Scan, preview, import, and install SKILL.md capabilities from one central library across supported agent platforms and project-local discovery.
Themes
Ten themes for the way you want to work
Change the entire application surface without changing the workflow. Choose from five focused dark themes and five expressive light themes, each applied consistently across Kanban, Swarm, Mux, and Skill modes.
Dark theme
Terminal CLI
High-contrast green phosphor styling for terminal-first focus.
Dark theme
Modern Dark
Quiet panels and softened geometry for long building sessions.
Light theme
Material You
Bright rounded surfaces with an accessible purple hierarchy.
Dark theme
Art Deco
Presentation-ready black and gold with crisp geometric framing.
Dark collection
Focused contrast for terminals, graphs, and long-running work.
- Terminal CLI
- Modern Dark
- Minimal Dark
- Cyberpunk
- Art Deco
Light collection
Clear daylight surfaces ranging from strict to approachable.
- Monochrome
- Swiss Minimalist
- Neo Brutalism
- Enterprise
- Material You
Tool Transparency
See what every coding-agent tool call is doing
Dorchestrator turns agent activity into concise queued, running, completed, and failed events. See which tool is called, why it is used, and the high-signal result without reading raw JSON or reconstructing progress from terminal output.
Queued
Read file context
Open source files needed to inspect Kanban timeline rendering.
Running
Search workspace
Find references to built-in agent tool event parsing.
Done
Patch renderer
Added bounded previews for tool summary and result details.
Failed
Run command
Command failed with permission error. User approval required.
Before
Users saw coarse status changes and generic queued signals. Reviewers had to infer progress from raw terminal output and final replies.
After
Each tool call appears as a stable row with action summary, lifecycle state, concise result preview, and visible failure context.
Why it matters
Transparency reduces guesswork, keeps long-running work understandable, and makes it easier to trust an agent while staying in the loop.
Local-First Design
Use your own AI providers from one desktop control layer
Dorchestrator coordinates the coding tools you configure; it does not resell model inference. Task boards, agent definitions, run history, templates, schedules, and related runtime records stay in the selected local workspace.
Built for independent software builders
Solo developers, indie hackers, technical founders, and AI-heavy builders use one workspace to coordinate several coding agents without adding a hosted control plane.
Supervise parallel agent tasks
Track long-running work on a Kanban board and in agent timelines instead of inferring progress across several raw terminal windows.
Keep workflow state with the project
Reusable templates, schedules, task state, agent definitions, and run history live with the selected workspace so recurring work is easier to restart.
Choose a desktop license
Choose Pro for continuous updates or Lifetime for a durable desktop license. Models stay on your own provider accounts, so the app never resells inference.
Bring Claude Code, Codex, or built-in agents
Use the coding tools and provider accounts you already have, then point each agent session at the same selected project workspace.
Review before accepting output
Final responses, transcripts, lifecycle events, and failure context stay attached to the work so you can accept, revise, or discard agent output.
Early Access
Join the Dorchestrator early-access waitlist
Dorchestrator is still in development. Join for launch updates, invitations to test the local AI coding-agent workspace, and a clear notice before paid plans open.
- Launch notice without feed noise
- Early workflow previews
- One-click unsubscribe from every update
Plans
Dorchestrator pricing for independent developers
Choose a subscription for continuous updates, or own the desktop license with a one-time purchase. Your model usage stays on your provider accounts.
$19 per user/month
For daily builders who want every workspace mode and continuous product updates.
- Kanban, Swarm, Mux, and Skill modes
- Tool timelines, run history, and review gates
- Scheduled commands and reusable templates
- Updates included while subscribed
$149
For builders who prefer one payment and a desktop version they can keep.
- All current Pro desktop features
- Personal license with one-time payment
- One year of product updates included
- Keep your last eligible version; renew updates optionally
Workflow
Plan, repeat, and review AI coding-agent work
Start
Start from the tools you already use: Claude Code, Codex, or the built-in coding agent, all pointed at the same local workspace.
Repeat
Convert repeated prompts into templates and scheduled automations so common coding loops become faster to restart.
Trust
Use transparent tool events and run history to see what happened before you accept, revise, or discard the agent output.
Frequently Asked Questions
Dorchestrator questions, answered
Direct answers about the desktop app, supported coding agents, local data, workspace modes, provider accounts, and operating systems.
What is Dorchestrator?
Dorchestrator is a local-first desktop application for orchestrating AI coding agents. It brings Claude Code, OpenAI Codex, and built-in agents into one workspace for task planning, multi-agent coordination, terminal sessions, reusable Skills, and run review.
Which AI coding agents does Dorchestrator support?
Dorchestrator can launch and coordinate Claude Code, OpenAI Codex, and its built-in coding agent. Mux templates can also run shell sessions and configured command-line tools in the selected workspace.
What do Kanban, Swarm, Mux, and Skill modes do?
Kanban delegates and reviews tasks, Swarm coordinates connected agents, Mux runs reusable multi-terminal layouts, and Skill manages reusable SKILL.md capabilities across supported agent platforms.
Is Dorchestrator local-first?
Yes. Workspace configuration, task boards, agent definitions, run history, templates, schedules, and related runtime records are stored locally, primarily under .dorchestrator in the selected project. Data can still be sent to a model provider when you direct an agent to use that provider.
Do I need separate Claude or OpenAI access?
Yes. You bring the provider accounts and credentials required by the coding tools you choose. Dorchestrator coordinates those tools, while third-party model usage, fees, and limits stay with your provider accounts.
Which operating systems does Dorchestrator support?
Dorchestrator is built for macOS, Windows, and Linux as a desktop developer tool.
Field Notes
Build better AI-agent workflows
Read practical guides, architecture notes, and lessons from building local-first coding-agent orchestration.