How it worksFeaturesPricingFAQGuidesMy account ESEN Get doki →

doki

Guides for running your coding agents

Plain pages on what each agent actually does, and how doki brings them into one Inbox. Useful when you are choosing a tool, and when several are already running.

Interactive tool

Which agent fits your work?

What do you want from the agent?

Where do you want to review the work?

How many sessions at once?

doki next to each agent

doki does not replace the agent that writes the code. These pages say what each one does, and when to run them together.

vsComparisonClaude Code vs dokiClaude Code writes code in the terminal. doki gathers those sessions into a Mac Inbox so you can answer with a key, by voice, or from your iPhone.▶ Interactive simulationvsComparisonCodex vs dokiOpenAI’s Codex CLI writes in the repo. doki shows those sessions next to Claude and Cursor, and replies in the terminal or by resuming the session.▶ Interactive simulationvsComparisonCursor vs dokiCursor is the editor and its agents. doki shows the app’s chats, the CLI, and Cursor Cloud, and can message them through the API without opening Cursor.▶ Interactive simulationvsComparisonGitHub Copilot vs dokiCopilot suggests and chats inside the editor. doki does not control it: doki steers terminal agents such as Claude Code, Codex, and Cursor. You can use both.▶ Interactive toolvsComparisonGemini CLI vs dokiGemini CLI is Google’s terminal agent. doki detects it by process and replies in that terminal, inside the same Inbox as the others.▶ Interactive simulationvsComparisonAider vs dokiAider pair-programs in the terminal and git. doki detects it by process, shows when it asks, and types the reply into that terminal.▶ Interactive simulationvsComparisonOpenCode vs dokiOpenCode is an open-source coding agent in the terminal. doki detects it by process and adds it to the Inbox next to Claude, Codex, and Cursor.▶ Interactive simulationvsComparisonFactory Droid vs dokiFactory Droid runs tasks in sessions under ~/.factory/sessions. doki shows them in the Inbox and replies with droid exec or in the terminal.▶ Interactive simulation

Which agent to use

Claude Code, Cursor, Codex and Copilot do different jobs. Here is the difference, and how to keep more than one open.

vsWhich agent to useClaude Code vs CursorClaude Code works in the terminal. Cursor is the editor and its agents. You do not have to pick one forever: doki attends to both in the same Inbox.▶ Interactive toolvsWhich agent to useClaude Code vs CodexClaude Code (Anthropic) and Codex CLI (OpenAI) are terminal agents. They differ by model, on-disk session, and how doki resumes each one.▶ Live reviewvsWhich agent to useCursor vs CodexCursor is an editor with agents in the app, the CLI, and the cloud. Codex is OpenAI’s terminal agent. doki shows both in one Inbox.▶ Interactive toolvsWhich agent to useClaude Code vs GitHub CopilotCopilot suggests inside the editor. Claude Code works on its own in the terminal until it asks. doki steers Claude, not Copilot chat.▶ Interactive tool

Working with several agents

Inbox, voice, iPhone, remote machines, permissions, worktrees and diff review. What doki does on a normal day.

GuideSeveral coding agents at onceHow to run Claude Code, Codex, and Cursor in parallel without losing their questions: an Inbox, worktrees, combos, and one place to answer.▶ Before and afterGuideAn inbox for coding agentsdoki’s Inbox gathers questions and finished work from Claude, Codex, and Cursor. You answer with J and K, a number, your voice, or a notification.▶ Interactive InboxGuideAnswer your agents from an iPhonedoki’s iPhone app shows the same Inbox as the Mac: options, approvals, photos, and voice notes, over the local network, a direct link, or the server.▶ iPhone simulationGuideCoding agents on other machinesdoki reaches a Mac mini, Linux box, or VPS over SSH with no open ports. With tmux you type live. doki-bridge covers other networks, with an optional relay.▶ Network diagramGuidePermissions for coding agentsGuard, in doki, classifies what an agent wants to run: allow, ask, or block. It explains the risk, and you approve on the Mac or the iPhone.▶ Rule testerGuideWorktrees so agents do not collidedoki gives each agent its own worktree: an isolated branch, a reserved port, and a copy of .env. When it finishes you merge, open a PR, or discard.▶ Live worktreesGuideSteer agents by voiceWith a double-tap of ⌥, doki transcribes what you say and sends it to the agent already on that task, or starts a new one. You can mark a region of the screen.▶ Voice simulationGuideReview the diff when the agent finishesWhen an agent finishes, doki reviews the diff and returns a verdict: ready, review, or don’t merge, with risks and secrets. It can send the findings back.▶ Live review

Never leave an agent waiting.

doki pulls Claude Code, Codex, Cursor and the other agents you already use into one Inbox, on your Mac, your iPhone and your servers.

See pricing →