iOS app

The Replicas iOS app is included with the Developer, Team, and Enterprise plans and is available to invited users through TestFlight. Accept the invitation on your iPhone, then open the latest available build. Sign in with email and password, Google, or Continue with Apple. Links to the Terms and Privacy Policy sit below the sign-in controls. New accounts complete four screens: Source Control, Repositories, Agent, and Integrations. Repository selection and integrations are optional; team invitations live in the sidebar checklist after setup. The app opens setup in a secure Safari view, then checks your progress when you return. CLI-based agent connections still require a computer. Use the centered organization picker at the top of the sidebar to switch organizations; that selection remains the default the next time you open the app. Tap your profile photo for account, photo, name, theme, support, and sign-out controls, or tap search on the right to filter workspaces. Environments use the same folder presentation as the web app. Tap + beside an environment to create a workspace directly. To start with a task, tap the prompt field at the bottom of the workspace list. Choose the environment at the top of the expanded composer and the harness and model at the bottom, then type your prompt or tap the microphone to dictate it. Tap the microphone again to stop recognition, then tap send. Typing while dictating stops recognition and keeps what you typed. Enter or Return adds a new line; use the send button to submit the prompt. Before sending, tap + to attach images, or copy an image and tap the paste control beside +. You can add up to five images of 20 MB each, including HEIC and HEIF photos. The iOS photo picker converts these photos to JPEG before sending. You can also enable Plan Mode, Codex Goal mode, or Fast Mode when supported by the selected agent. Replicas creates the workspace in the selected environment and sends the prompt, images, and selected modes when it is ready. New workspaces open in chat while they prepare. You can queue the first message immediately, and Replicas sends it when provisioning finishes. After the first successful message, Replicas names the workspace from your prompt. Expand an environment to show each workspace once. Long-press an environment or workspace, then drag it to reorder the list. The lifted item follows your finger while nearby items make room, then springs into place when released. Swipe a workspace left across the row to archive it, or release a short swipe to spring back. Pin a workspace to move it into the dedicated Pinned folder at the top; unpinning returns it to its environment. A green dot marks an unread agent reply until you open the workspace, and the Replicas loader replaces the trailing controls while a workspace is loading or processing. Rows with linked pull requests use the same color-coded state icon, count, and proportional mixed-state indicator as the web sidebar. Agent activity is grouped into a collapsible timeline for each turn. It stays open and shows live read, command, search, and edit steps while the agent works; when the turn finishes, collapse or reopen it to see the completed steps and changed-file summary. Mobile does not expose repository or inline patch diff viewers. Workspace data is cached separately for each organization so it appears immediately when you reopen the app, then refreshes when the app returns to the foreground. Pinned workspaces, unread state, notes, drafts, ordering, and interface preferences also sync with the web and desktop apps. The app shows its local cached copy immediately, then refreshes it from your account and reconciles any differences.

​

Chat with an agent

Open a workspace to follow streaming agent replies, reasoning, commands, file changes, searches, and plan progress. Swipe right from the left edge of the chat to return to the workspace list. Agent replies render Markdown, including tables that scroll horizontally. Tap an activity to expand its details. When you reopen a workspace, the app selects its most recently used started chat and restores that chat’s harness and model context. Tap the glass ellipsis button in the workspace header to switch between independent conversations, or tap + in the chat list to start another with a different coding agent. Opening a long chat shows loading feedback while its transcript loads, then starts at the newest messages. New streaming output smoothly follows the latest message while you remain near the bottom. When you scroll back through a long chat, automatic following pauses; tap the down arrow to return to the latest message. At the top of the conversation, the header stays clear of the first message. On iOS 26, the elevated chat header, the down arrow, and the View PR action use native Liquid Glass; earlier iOS versions use a translucent fallback. When a workspace has linked GitHub pull requests in a stack, the chat header lists them in stack order. Tap a pull request to review or merge it on GitHub; direct stack merging is unavailable in Replicas because stacks must be merged in order. For a standalone pull request, tap View PR above the message field to open it; on iOS 26 it uses the same Liquid Glass material as the chat header. Sent messages appear immediately, while chat history, queues, and agent status stay synchronized as background work progresses. A conversation or its earlier history that fails to load shows the error with a Retry action. Messages originating from Slack, Linear, GitHub, GitLab, automations, and CI appear as source-specific cards. Tap linked metadata such as an issue, pull request, workflow run, or sender to open its source, and expand context sections to inspect threads, review diffs, or failure details without leaving the conversation. The message field and send or stop control share one compact row on an opaque panel that floats above the transcript. Tap the microphone to dictate into the prompt, then tap it again to stop recognition; typing while dictating also stops it and keeps what you typed. Tap send separately to submit the transcribed text. Tap the field to expand the composer, which morphs into its taller shape, and reveal the harness, model, Plan, Fast, Goal, and + controls supported by the selected agent. Tap a mode to toggle it for the next message. Tap + to manage the same modes alongside context usage, photo, video, and file actions in a bottom sheet, or touch and hold the message field to paste text. Copy an image and tap the paste control beside + to attach it, with the same supported formats and five-image and 20 MB limits as a new task. The iOS photo picker converts HEIC and HEIF photos to JPEG before sending. Photos remain inline attachments; videos up to 25 MB upload to workspace Canvas as file attachments. The organization picker uses the same presentation. New chats use your saved harness and model preferences. Tap the harness and model label to switch harnesses, choose a compatible model, or change the effort level; switching harnesses updates the available models automatically. Attached items appear above the expanded composer and can be removed before sending. The composer and message list move with the keyboard as it opens and closes, and settle without animation when Reduce Motion is on. Text and inputs follow your iPhone text-size and accessibility settings. You can send a follow-up while an agent is working, but the mobile app does not expose queue-management controls. The workspace menu provides sharing and workspace details. Files, Canvas, Media, Code, and Desktop surfaces are not available in the mobile app.

​

AI data sharing

Before the first agent run, the app asks you to allow sending your prompt, selected attachments, and relevant workspace context to the AI provider you choose. Sending a message, creating a workspace from a task, or answering an interactive agent question stays pending until you allow it; Not now leaves the request unsent. Review or withdraw the permission any time from your profile under AI data sharing. Withdrawing it does not affect data already processed.

​

Support and account controls

Your profile menu links to support, the Privacy Policy, and the Terms of Service, and includes Report agent output for flagging a response by email. Tap Delete account to request permanent deletion of your account and associated personal data without contacting support. The app confirms before submitting, then shows the date deletion completes by. Tap Cancel deletion request before processing begins to keep the account and its data. You receive an email when deletion is complete. To install a team build, accept the TestFlight invitation on your iPhone and open the latest available version.