# Starkomand: AI agent orchestration for coding on Mac AI agent orchestration brings your coding agents and their work onto one canvas. Starkomand runs Claude Code, Codex, Cursor and other configured CLI agents in their own terminal windows on an infinite canvas. Drop a GitHub issue on an agent, ask an agent to draw the architecture beside its terminal, hand the work to another agent with a focused brief, and hold Space to flip every window to a short recap. The homepage opens with a live demo of those four moves: the real app playing a scripted scenario in the browser, followed by one scene per move. The demo is scripted; the application does the same with your own agents, repository and provider sign-ins. The objective, acceptance criteria, handoff briefs and review evidence stay with the task as work moves between agents. One agent is enough to start; handoffs and review loops are optional. The application is for macOS on Apple silicon. The current offering does not include Windows, Linux or Intel Mac downloads. Use the [homepage](https://starkomand.com/) for the current Mac download and [release notes](https://github.com/developertimeline/starkomand-releases/releases) for published changes. ## Supported agents Claude Code (`claude`), Codex (`codex`), Cursor (`cursor-agent`), Gemini CLI (`gemini`), Grok CLI (`grok`), GitHub Copilot (`copilot`), OpenCode (`opencode`, including GLM via Z.ai), Amp (`amp`), Pi (`pi`), Aider (`aider`) and Goose (`goose`), plus any other CLI agent you configure. Each runs signed in as you, as in any terminal. Starkomand does not resell tokens or ask for API keys, and you can open several sessions of the same provider. ## Getting started 1. Download Starkomand and choose your project folder in a workspace. 2. Launch one installed coding agent using your existing provider sign-in. 3. Ask it to explain the project, propose a small improvement, identify relevant files and write acceptance criteria. Request a plan only if you are not ready for edits. 4. Inspect the answer. Continue with the same agent or use Workflows to prepare a focused handoff. Choose implementation explicitly when you want changes, then review the result. Starkomand organizes agent sessions and their context. You can continue using your editor. AI subscriptions and usage are billed separately by the providers. ## Multi-agent orchestration on the canvas - Save the task objective, acceptance criteria, evidence and handoff history across agent sessions. - Choose a model and stage for a focused handoff to a fresh agent. - Run a single review or opt into bounded fix-and-review rounds with final QA. - Pause a review loop when repository changes make its review evidence stale. - Arrange real agent terminals on an infinite canvas. - Place a GitHub issue on the canvas and drop it onto an agent window to hand over the issue; a note dropped on a busy agent waits its turn. - Ask an agent to draw an editable architecture diagram beside the code. - Open browser previews alongside the agent working on the project. - Hold Space to see agent recaps, and use Tidy to seat a handoff run together. - Use GitHub tools to see pull requests that need attention. - Keep separate workspaces with their own agents, folder, layout and backdrop. - Customize the canvas with included backdrops or a local image. Hiding the canvas keeps agents working while the local server runs. Restarting the Mac ends live sessions but preserves the canvas, notes and task history. Review evidence records what was checked against a particular revision. An agent verdict does not guarantee correctness or accept the task; the engineer decides when work is complete. See the [homepage scenes](https://starkomand.com/#workflow) for captures of the app delivering an issue, drawing a diagram and taking voice commands, a still of a handoff run with an example brief for a pagination issue, and an illustrative usage-limit recovery. ## Voice control for coding agents and optional gestures Type, drag and use keyboard shortcuts as usual. Voice control and hand gestures are optional. Hold Option to speak to an agent by name, open agents, group a squad, tidy windows or switch workspaces. Local dictation requires an optional download. Local dictation and hand tracking process audio and camera frames on the Mac. Optional cloud dictation sends audio to OpenAI, and text sent to an agent goes to its provider. ## Trial and Pro The 3-day trial includes all features. Continuing after the trial requires Starkomand Pro, a one-time purchase for one Mac that includes a year of updates. You can keep using the purchased version after that year. Pro and the trial include unlimited agents, workspaces and Command Keys; supported providers; voice, gestures and squads; diagrams, recaps, previews and GitHub tools; and all backdrops, including your own images. Launch prices change as seats sell. Check the [pricing section](https://starkomand.com/#pricing) and checkout for the current offer and availability. Agent subscriptions and usage are separate from the Starkomand license. ## Privacy There is no Starkomand account. Workspace data, local dictation, hand tracking and recaps stay on the Mac by default. Agents and connected features use their providers with your sign-ins. Remote recaps send terminal context only after you explicitly select Claude or Codex in Privacy settings, with no cross-provider fallback. Trial verification and licensing use their described services; workspace content is not included in those requests. See the [privacy policy](https://starkomand.com/privacy/) for data flows and the [terms and conditions](https://starkomand.com/terms/) for licensing and refund terms. ## Who makes it Starkomand is a [Developer Timeline](https://developertimeline.io) product, built by Rashid, a software engineer of eight years. Developer Timeline has 128K followers on [Instagram](https://www.instagram.com/developertimeline) and 19K subscribers on [YouTube](https://www.youtube.com/@developertimeline). ## Compared with other tools Starkomand keeps the real agent CLI in a real terminal window and adds the canvas and tools around it; it runs on your Mac with no account and a one-time license. The [comparison page](https://starkomand.com/compare/) sets it beside Conductor, Superset, Emdash, Maestri, AgentGrid, CNVS, 49Agents, Zeron and hosted coding agents such as Devin and Cursor Cloud Agents, with each product's price, platform, supported agents and where the code runs, taken from their public pages on 28 September 2026 (Superset and Emdash on 9 October 2026). One page per rival shows only the rows where the two differ and says which to pick: [Starkomand vs AgentGrid](https://starkomand.com/compare/agentgrid/), [Starkomand vs Maestri](https://starkomand.com/compare/maestri/), [Starkomand vs Conductor](https://starkomand.com/compare/conductor/), [Starkomand vs Superset](https://starkomand.com/compare/superset/), [Starkomand vs Emdash](https://starkomand.com/compare/emdash/), [Starkomand vs CNVS](https://starkomand.com/compare/cnvs/), [Starkomand vs 49Agents](https://starkomand.com/compare/49agents/) and [Starkomand vs Zeron](https://starkomand.com/compare/zeron/). For someone who has not chosen a tool yet, an alternatives page per product lists the other apps with their price, platform and who each suits ([Conductor alternatives](https://starkomand.com/compare/conductor-alternatives/), [Superset alternatives](https://starkomand.com/compare/superset-alternatives/), [Emdash alternatives](https://starkomand.com/compare/emdash-alternatives/), [Maestri alternatives](https://starkomand.com/compare/maestri-alternatives/), [AgentGrid alternatives](https://starkomand.com/compare/agentgrid-alternatives/), [CNVS alternatives](https://starkomand.com/compare/cnvs-alternatives/), [49Agents alternatives](https://starkomand.com/compare/49agents-alternatives/), [Zeron alternatives](https://starkomand.com/compare/zeron-alternatives/)), and [Run multiple Claude Code agents in parallel](https://starkomand.com/compare/run-multiple-claude-code-agents/) walks the git worktree setup by hand before comparing the apps. ## Use cases The same canvas, set up for different work. [Freelance developers](https://starkomand.com/for/freelance-developers/) keep one workspace per client, with its own folder and agents. [Software agencies](https://starkomand.com/for/agencies/) hand client issues to agents and have a second agent review the work; licenses are per Mac, with no shared team canvas. [Enterprise developers](https://starkomand.com/for/enterprise-developers/) map a large codebase, work their tickets in parallel and get a second agent’s review before opening a pull request, with the agent CLIs and sign-ins their company already uses. [Indie hackers](https://starkomand.com/for/indie-hackers/) run a crew that plans, builds and reviews. [SaaS founders](https://starkomand.com/for/saas-founders/) drop customer bugs on agents and keep the pull requests that need them in view. [Open source maintainers](https://starkomand.com/for/open-source-maintainers/) work through GitHub issues with agents. The [use-case hub](https://starkomand.com/for/) lists them all. ## Support - [Frequently asked questions](https://starkomand.com/#faq): Compatibility, supported agents, resource use, sessions and licensing. - [Find your license key](https://starkomand.com/license/): Enter your purchase email to request a recovery link from Dodo Payments. - [Agent overview](https://starkomand.com/llms.txt): Concise product summary and links to official pages.