dog.codes

Every coding agent. One workspace. Command it from anywhere.

English · 简体中文 · 繁體中文 · 日本語 · 한국어 · Español · Français · Deutsch · Português (Brasil)

dog.codes brings your coding agents, conversations, code changes, and project commands into one desktop app. Let one agent implement a feature, ask another to review it, then run the project and prepare a commit—all from the same workspace. When you step away from your desk, command it all from your phone.

Website · Features · Phone remote · Get started

Supported agents

Connect the agents you already use. The 12 integrations below are implemented; available models, approvals, usage data, and recovery behavior vary by agent and installed version. Install and sign in to your chosen agent through its own tools first.

One place to work

Features

Hand off work between agents seamlessly

Switch the agent running the same conversation at any time; context is handed over automatically.

• Give each task to the agent best suited for it — one designs the architecture, another fills in the details.
• When one agent runs out of quota, another picks up the work without you restating the requirements.
• The previous conversation travels with the handoff, and the new agent can look up earlier turns through built-in history tools — it truly continues instead of starting over.

Know your quota and usage at a glance

Track quotas, usage, reset times, and versions for all 12 supported coding agents in one place.

• See every agent's remaining quota and reset time, plan your work around them, and make every token count.
• Manage agent versions in one place and upgrade with a single click.
• Installed agents are discovered automatically, along with their sign-in status.

Cross-agent code review and repair

Ask a different agent to review the code that was just changed, catching bugs and design flaws, or to run a security or performance analysis.

• Different agents and models review each other's work from a fresh context, avoiding the blind spots of a single agent in one conversation.
• Findings are ranked high, medium, and low; select the ones you want and start the repair with one click.
• The repair goes back to the main agent together with the review report and selected findings, keeping the context intact.

Fork across agents and compare approaches

Fork a new conversation from any turn and let another agent try a different idea, leaving the original untouched.

• With worktrees, the code changes of each approach stay isolated from one another.
• Let different agents each produce a version of the same task, then keep the best one.

Git management that feels familiar

Git operations modeled closely on VS Code, so you manage the Diff in the sidebar the way you already do.

• Stage, unstage, discard, ignore, Commit, and Commit & Push, with tree and list views.
• Have any agent write the commit message with one click, in the language you choose.

Discover how to run a project, then run it in one click

dog.codes analyzes how a project is run and debugged and turns those commands and scripts into clickable launch entries.

• No more remembering each project's start command or opening a terminal just to launch it — one click runs it.
• Output is collected in one place so you can check it any time.

Multi-folder projects

Frontend, backend, and docs in separate folders? Hand them all to the agent in a single conversation.

• Even agents without native multi-folder support can work on multi-folder projects through dog.codes.

Per-turn usage, crystal clear

See the token usage, context occupancy, cache hits, and cache usage of every turn.

• Understand exactly what the current conversation and each turn consumed, and where your tokens went.

Run agents on your remote hosts

Keep your code and agents on your own servers. Connect over SSH and manage remote development from the same dog.codes workspace.

• Install supported agent CLIs on the host in one click, then manage installed versions, updates, remaining quotas and reset times in the UI.
• Browse directories on the host visually and open a folder as a project, without copying the repository to your computer.
• Choose an agent running on that host, assign tasks, and follow streamed responses, permission requests and questions in the familiar conversation UI.
• Work with remote files, Git changes, code reviews, terminals and project run commands from the desktop.

Codex-level computer use for every agent

Give agents and CLIs that cannot operate your computer or verify their own work the same powerful computer-use capabilities as Codex.

This is the beginning of the README. Read the rest at the original. Read the full README ↗