[{"data":1,"prerenderedAt":36},["ShallowReactive",2],{"doc:en:quick-start":3},{"title":4,"lede":5,"html":6,"headings":7},"Quick start","A first session step by step: add a project, create an agent, choose its model, effort and permission mode, give it a request, answer its questions, review what it changed and commit.","\u003Ch2 id=\"before-you-start\">Before you start\u003C\u002Fh2>\n\u003Cp>You need Claude Code installed and signed in to your Claude account, and git (see \u003Ca href=\"\u002Fen\u002Fdocs\u002Finstallation\u002F\">installation\u003C\u002Fa>). On macOS, \u003Ccode class=\"key\">Ctrl\u003C\u002Fcode> becomes \u003Ccode class=\"key\">⌘\u003C\u002Fcode> in the shortcuts of this page.\u003C\u002Fp>\n\u003Ch2 id=\"add-a-project\">Add a project\u003C\u002Fh2>\n\u003Cp>A project is a folder of code. On the welcome screen, click “+ Add a project”. Later, the \u003Ccode class=\"key\">+\u003C\u002Fcode> button to the right of the tabs in the title bar does the same. The “New project” window opens:\u003C\u002Fp>\n\u003Cdiv class=\"table\">\u003Ctable>\n\u003Cthead>\n\u003Ctr>\n\u003Cth>Field\u003C\u002Fth>\n\u003Cth>What it does\u003C\u002Fth>\n\u003C\u002Ftr>\n\u003C\u002Fthead>\n\u003Ctbody>\u003Ctr>\n\u003Ctd>“Folder”\u003C\u002Ftd>\n\u003Ctd>The path of the project. “Browse…” opens the folder picker. Under the field, a line says what Escouade knows about the folder: “Git repository detected · branch main · clean” (or the number of changed files), “No git repository · one will be initialized” or “Folder not found”.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>“Name”\u003C\u002Ftd>\n\u003Ctd>Filled in with the folder’s name, which you can change. “Tab preview” shows the tab as it will look.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>“Color”\u003C\u002Ftd>\n\u003Ctd>One of the 14 colors offered; the first one no project uses yet is picked for you. It tints the tab and the interface while the project is on screen.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>“Create a first agent”\u003C\u002Ftd>\n\u003Ctd>On by default: “Opens a conversation in this project right away”. The “Fable”, “Opus”, “Sonnet” and “Haiku” buttons, each with its version, choose its model; the one of the “Default model” setting is already selected.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>“One git worktree per agent”\u003C\u002Ftd>\n\u003Ctd>Off by default: “Keeps each agent’s work apart and shows its changed files separately”. You can turn it on later, see \u003Ca href=\"\u002Fen\u002Fdocs\u002Fworktrees\u002F\">worktrees\u003C\u002Fa>.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003C\u002Ftbody>\u003C\u002Ftable>\u003C\u002Fdiv>\n\u003Cp>Click “Create project” (the button stays grayed out until the folder exists). The project opens in a new tab, with its first agent if you asked for one.\u003C\u002Fp>\n\u003Cdiv class=\"callout note\">\u003Cp class=\"callout-title\">Note\u003C\u002Fp>\u003Cp>If the folder is not inside a git repository, Escouade runs \u003Ccode>git init\u003C\u002Fcode> to create one: without git, there would be no diff and no commit to review.\u003C\u002Fp>\n\u003C\u002Fdiv>\n\u003Ch2 id=\"create-an-agent\">Create an agent\u003C\u002Fh2>\n\u003Cp>An agent is a conversation with Claude Code in this project. To create one, press \u003Ccode class=\"key\">Ctrl+N\u003C\u002Fcode> or click “+ New agent” at the top of the sidebar: the agent is created without asking you anything, and its conversation appears. It is called \u003Ccode class=\"key\">agent-1\u003C\u002Fcode>, \u003Ccode class=\"key\">agent-2\u003C\u002Fcode>… until its first request; Escouade then gives it a short name taken from what you wrote. To rename it yourself, double-click its name.\u003C\u002Fp>\n\u003Cp>Its status is “Ready”, and the empty conversation says “Agent ready”, with the folder it works in. The ▾ arrow to the right of “+ New agent” offers “New agent on a branch…” (see \u003Ca href=\"\u002Fen\u002Fdocs\u002Fbranches\u002F\">branches\u003C\u002Fa>).\u003C\u002Fp>\n\u003Ch2 id=\"choose-the-model-the-effort-and-the-permission-mode\">Choose the model, the effort and the permission mode\u003C\u002Fh2>\n\u003Cp>Under the message field, three menus set up the agent (a fourth one, “Account”, appears when you have several Claude accounts). You can change them in the middle of a conversation.\u003C\u002Fp>\n\u003Cdiv class=\"table\">\u003Ctable>\n\u003Cthead>\n\u003Ctr>\n\u003Cth>Menu\u003C\u002Fth>\n\u003Cth>Choices\u003C\u002Fth>\n\u003C\u002Ftr>\n\u003C\u002Fthead>\n\u003Ctbody>\u003Ctr>\n\u003Ctd>“Model”\u003C\u002Ftd>\n\u003Ctd>“Fable”, “Opus”, “Sonnet” and “Haiku”. Each runs the latest model of its family that Claude Code knows, and shows its version (for example “Sonnet 5.5”).\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>“Effort”\u003C\u002Ftd>\n\u003Ctd>“Low” (quick answers, little thinking), “Medium” (balanced), “High” (deep thinking), “Very high” (very deep thinking), “Max” (maximum thinking, uses more tokens). Haiku doesn’t support effort.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>“Mode”\u003C\u002Ftd>\n\u003Ctd>The permission mode, described below.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003C\u002Ftbody>\u003C\u002Ftable>\u003C\u002Fdiv>\n\u003Cp>The permission modes decide when Claude asks you for permission to act:\u003C\u002Fp>\n\u003Cdiv class=\"table\">\u003Ctable>\n\u003Cthead>\n\u003Ctr>\n\u003Cth>Mode\u003C\u002Fth>\n\u003Cth>What happens\u003C\u002Fth>\n\u003C\u002Ftr>\n\u003C\u002Fthead>\n\u003Ctbody>\u003Ctr>\n\u003Ctd>“Auto”\u003C\u002Ftd>\n\u003Ctd>A classifier approves safe actions and asks about the rest. Unavailable with Haiku.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>“Ask”\u003C\u002Ftd>\n\u003Ctd>Claude asks for your approval before each sensitive action.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>“Plan”\u003C\u002Ftd>\n\u003Ctd>Claude analyzes and proposes a plan without changing anything.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>“Accept edits”\u003C\u002Ftd>\n\u003Ctd>File edits are accepted without asking.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>“Bypass”\u003C\u002Ftd>\n\u003Ctd>No permission requests.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003C\u002Ftbody>\u003C\u002Ftable>\u003C\u002Fdiv>\n\u003Cp>A new agent starts with the values of Settings › Claude Code: “Default model” (Sonnet), “Default effort” (Medium) and “Default permission mode” (Auto). See the \u003Ca href=\"\u002Fen\u002Fdocs\u002Fsettings-claude-code\u002F\">Claude Code settings\u003C\u002Fa> and \u003Ca href=\"\u002Fen\u002Fdocs\u002Fmanaging-agents\u002F\">agents\u003C\u002Fa>.\u003C\u002Fp>\n\u003Cdiv class=\"callout warning\">\u003Cp class=\"callout-title\">Warning\u003C\u002Fp>\u003Cp>In “Bypass” mode, Claude acts without any permission request: Escouade reminds you of it when you choose it. Keep it for safe environments.\u003C\u002Fp>\n\u003C\u002Fdiv>\n\u003Ch2 id=\"send-a-first-request\">Send a first request\u003C\u002Fh2>\n\u003Cp>Click in the message field, describe the task (“Describe the task to give to Claude…”), then press \u003Ccode class=\"key\">Enter\u003C\u002Fcode> to send. \u003Ccode class=\"key\">Shift+Enter\u003C\u002Fcode> starts a new line.\u003C\u002Fp>\n\u003Cul>\n\u003Cli>Type \u003Ccode class=\"key\">@\u003C\u002Fcode> followed by a bit of a file name to pick a file of the project, or \u003Ccode class=\"key\">\u002F\u003C\u002Fcode> to pick a command.\u003C\u002Fli>\n\u003Cli>Paste or drop an image, a PDF or a text file into the field, or click the paperclip, to attach it to the message.\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>As soon as you send, the agent’s status changes to “Running” and the conversation says “Claude is working…”. Its tool calls show up as compact rows that you can unfold. You can write to it while it works: it takes your message into account at its next step. To stop it, press \u003Ccode class=\"key\">Esc\u003C\u002Fcode> in the message field or click “■ Stop”.\u003C\u002Fp>\n\u003Ch2 id=\"answer-a-question-or-a-permission-request\">Answer a question or a permission request\u003C\u002Fh2>\n\u003Cp>When Claude needs you, a card with an amber border appears in the conversation, the agent’s status becomes “Question”, and its tab and card blink if you are looking elsewhere. There are two cards.\u003C\u002Fp>\n\u003Cp>\u003Cstrong>“Claude is asking for permission”\u003C\u002Fstrong> names the tool and what it is about to do. Three buttons:\u003C\u002Fp>\n\u003Cul>\n\u003Cli>“Allow” (\u003Ccode class=\"key\">Ctrl+Enter\u003C\u002Fcode>);\u003C\u002Fli>\n\u003Cli>“Always allow” (\u003Ccode class=\"key\">Ctrl+Shift+Enter\u003C\u002Fcode>), when Claude Code offers to remember the choice;\u003C\u002Fli>\n\u003Cli>“Deny”.\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>To deny and explain what to do instead, write it in the message field and send it. When Claude Code would deny the request by default, “Deny” is the main button and only a click on “Allow” grants it: the keys don’t answer. When Claude proposes a plan, the card is called “Claude proposes a plan” and the buttons are “Approve the plan”, “Approve and accept edits” and “Keep planning”.\u003C\u002Fp>\n\u003Cp>\u003Cstrong>“Claude is waiting for your answer”\u003C\u002Fstrong> asks a question with options. Click an option, or press \u003Ccode class=\"key\">Alt+1\u003C\u002Fcode> to \u003Ccode class=\"key\">Alt+9\u003C\u002Fcode>, or answer freely in the message field. With a single single-choice question, the answer goes out as soon as you click; with several questions or a multiple choice, “Submit” (\u003Ccode class=\"key\">Ctrl+Enter\u003C\u002Fcode>) sends the answers.\u003C\u002Fp>\n\u003Cdiv class=\"callout tip\">\u003Cp class=\"callout-title\">Tip\u003C\u002Fp>\u003Cp>\u003Ccode class=\"key\">Ctrl+J\u003C\u002Fcode> opens the next agent that is waiting for your answer, with the cursor in its message field. A card that has just arrived ignores the answer keys for half a second, so that a keystroke meant for your message doesn’t answer in its place.\u003C\u002Fp>\n\u003C\u002Fdiv>\n\u003Ch2 id=\"see-what-the-agent-changed\">See what the agent changed\u003C\u002Fh2>\n\u003Cp>When Claude has finished, the status becomes “Done” and a “Task done” card gives the duration, the tokens, the cost and the list of edited files; a click on a file opens it in the editor.\u003C\u002Fp>\n\u003Cp>To review the changes, open the “Uncommitted” panel: click “Files ▸” in the header of the conversation, or press \u003Ccode class=\"key\">Ctrl+Shift+B\u003C\u002Fcode>.\u003C\u002Fp>\n\u003Cul>\n\u003Cli>Each row is a file, with its letter (M for modified, A for added, D for deleted) and its added and removed lines.\u003C\u002Fli>\n\u003Cli>A click on a file opens its diff in a large window. “Unified” and “Side by side” change the layout, \u003Ccode class=\"key\">Esc\u003C\u002Fcode> closes the window. When there are several files, a column on the left lets you switch from one to another.\u003C\u002Fli>\n\u003Cli>“Show diff” opens the diff of all the files in the list.\u003C\u002Fli>\n\u003Cli>“This agent” lists the agent’s files; “Whole project” lists those of all the project’s agents.\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>To keep the conversation and the diff in view together, switch to the side-by-side layout with \u003Ccode class=\"key\">Ctrl+Shift+L\u003C\u002Fcode>: the panel stays displayed on the right and shows the diff of the chosen file under the list.\u003C\u002Fp>\n\u003Cp>\u003Cimg src=\"\u002Fimages\u002Fen\u002Fgit.jpg\" alt=\"An agent whose “Uncommitted” panel lists three files, with the diff of the first one to the right of the conversation\" width=\"1920\" height=\"960\" loading=\"lazy\" decoding=\"async\">\u003C\u002Fp>\n\u003Ch2 id=\"commit\">Commit\u003C\u002Fh2>\n\u003Cp>The “Commit…” button of the panel commits the agent’s changes (“Commit all…” when “Whole project” is chosen). What happens depends on the project’s “Commit” setting, in Settings › Projects:\u003C\u002Fp>\n\u003Cul>\n\u003Cli>\u003Cstrong>“Written by the agent”\u003C\u002Fstrong> (the default): Escouade sends the agent a message asking it to commit its changes with a clear message in Conventional Commits format. A short-lived message confirms “Commit request sent to” followed by the agent’s name, then the agent commits like it does for any request.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>“Direct, with a suggested message”\u003C\u002Fstrong>: the “Commit” window lists the files, Haiku writes a message that you read and edit, “Regenerate” asks for another one, and “Commit” commits. Nothing is committed without that click.\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>If the agent works in a worktree, the panel also offers “Merge … → …”, which merges its branch into the project’s. See \u003Ca href=\"\u002Fen\u002Fdocs\u002Fgit\u002F\">Git and commits\u003C\u002Fa> and \u003Ca href=\"\u002Fen\u002Fdocs\u002Fworktrees\u002F\">worktrees\u003C\u002Fa>.\u003C\u002Fp>\n\u003Ch2 id=\"what-next\">What next\u003C\u002Fh2>\n\u003Cul>\n\u003Cli>\u003Ca href=\"\u002Fen\u002Fdocs\u002Fmanaging-agents\u002F\">Agents\u003C\u002Fa> and \u003Ca href=\"\u002Fen\u002Fdocs\u002Fchat\u002F\">the conversation\u003C\u002Fa>: permission modes, messages, images, search, history.\u003C\u002Fli>\n\u003Cli>\u003Ca href=\"\u002Fen\u002Fdocs\u002Fnotifications\u002F\">Notifications\u003C\u002Fa> and \u003Ca href=\"\u002Fen\u002Fdocs\u002Foverview\u002F\">keeping track of agents\u003C\u002Fa>: get notified, answer without opening each conversation.\u003C\u002Fli>\n\u003Cli>\u003Ca href=\"\u002Fen\u002Fdocs\u002Fgit\u002F\">Git and commits\u003C\u002Fa>, \u003Ca href=\"\u002Fen\u002Fdocs\u002Fworktrees\u002F\">worktrees\u003C\u002Fa> and \u003Ca href=\"\u002Fen\u002Fdocs\u002Fbranches\u002F\">branches\u003C\u002Fa>: review, commit, isolate agents.\u003C\u002Fli>\n\u003Cli>\u003Ca href=\"\u002Fen\u002Fdocs\u002Feditor\u002F\">The editor\u003C\u002Fa>, \u003Ca href=\"\u002Fen\u002Fdocs\u002Fterminals\u002F\">terminals\u003C\u002Fa> and \u003Ca href=\"\u002Fen\u002Fdocs\u002Flaunch\u002F\">running the project\u003C\u002Fa>.\u003C\u002Fli>\n\u003Cli>\u003Ca href=\"\u002Fen\u002Fdocs\u002Fboard\u002F\">The Kanban\u003C\u002Fa> and \u003Ca href=\"\u002Fen\u002Fdocs\u002Fautopilot\u002F\">autopilot\u003C\u002Fa>: hand tickets to agents.\u003C\u002Fli>\n\u003Cli>\u003Ca href=\"\u002Fen\u002Fdocs\u002Faccounts\u002F\">Claude accounts\u003C\u002Fa> and \u003Ca href=\"\u002Fen\u002Fdocs\u002Fusage\u002F\">quotas and cost\u003C\u002Fa>.\u003C\u002Fli>\n\u003Cli>\u003Ca href=\"\u002Fen\u002Fdocs\u002Fshortcuts\u002F\">The keyboard shortcuts\u003C\u002Fa> and \u003Ca href=\"\u002Fen\u002Fdocs\u002Finterface\u002F\">the interface\u003C\u002Fa> for a tour of the window.\u003C\u002Fli>\n\u003C\u002Ful>\n",[8,12,15,18,21,24,27,30,33],{"id":9,"text":10,"depth":11},"before-you-start","Before you start",2,{"id":13,"text":14,"depth":11},"add-a-project","Add a project",{"id":16,"text":17,"depth":11},"create-an-agent","Create an agent",{"id":19,"text":20,"depth":11},"choose-the-model-the-effort-and-the-permission-mode","Choose the model, the effort and the permission mode",{"id":22,"text":23,"depth":11},"send-a-first-request","Send a first request",{"id":25,"text":26,"depth":11},"answer-a-question-or-a-permission-request","Answer a question or a permission request",{"id":28,"text":29,"depth":11},"see-what-the-agent-changed","See what the agent changed",{"id":31,"text":32,"depth":11},"commit","Commit",{"id":34,"text":35,"depth":11},"what-next","What next",1791926756101]