manaflow-ai/cmux

Features

• Coding Agent Orchestrators - Coordinates multiple parallel AI coding agents using a GPU-accelerated terminal with scriptable workspaces.
• Layout Organization - Translates orchestration commands into native window splits and terminal streams for agent organization.
• Teammate Pane Spawning - Programmatically spawns teammate AI agents into native UI split panes for coordinated work.
• Agent Team Orchestration - Provides orchestration and organizational management for teams of AI agents within terminal splits.
• Coding Agent Integrations - Integrates supported AI coding agents as dedicated terminal surfaces to coordinate development workflows.
• Multi-Agent Orchestrators - Coordinates parallel sub-agents by mapping their communication to native workspace panes and notifications.
• High-Density Workspace Managers - Organizes high-density coding sessions using vertical tabs, grouped environments, and real-time metadata.
• Editor Split Pane Controllers - Creates split-pane layouts and manages focus across multiple terminal surfaces within a workspace.
• Browser Automation Engines - Integrates a specialized browser automation engine within the terminal for DOM manipulation and accessibility tree capture.
• Headless Browser Automation - Provides a headless browser engine for programmatic DOM manipulation and session state capture.
• Integrated Web Browsing - Integrates a built-in web browser for rendering and navigating pages as buffers within the terminal environment.
• Parallel Session Tracking - Tracks multiple concurrent sessions in a sidebar showing git branches, directories, and live output.
• Socket-Based Orchestration APIs - Provides a high-level socket API for the programmatic orchestration of terminal behavior and AI agent workflows.
• Split Pane Layouts - Divides workspaces into multiple split regions to enable side-by-side viewing of concurrent AI agent sessions.
• Socket-Based Control Planes - Exposes a Unix domain socket and CLI for programmatic management of layouts and input delivery.
• Terminal Window Managers - Controls windows, workspaces, panes, and focus through a CLI to ensure deterministic placement.
• Webview Automation - Provides programmatic control of embedded webview surfaces using DOM actions and session state capture.
• Multi-Agent Sidebar Organization - Organizes concurrent sessions in a vertical sidebar with real-time metadata for git branches and directories.
• Terminal Workspace Layouts - Implements predefined arrangements of split terminals and browser panes with specific startup commands.
• Interface Density Scaling - Handles dozens of concurrent sessions using a high-density system of vertical tabs and split panes.
• Read-State Tracking - Tracks and cycles through read and unread statuses to help users prioritize attention across multiple AI agent outputs.
• Interface Relays - Relays terminal commands, HTTP traffic, and notifications over SSH to render remote activity locally.
• Unix Socket Interfaces - Exposes a Unix domain socket that accepts JSON requests for programmatic control of the application.
• GPU-Accelerated Renderers - Uses a GPU-accelerated rendering engine to ensure low-latency display updates and smooth scrolling of terminal text.
• Agent Execution Environments - Executes coding agents on remote machines while rendering the interface as native splits locally.
• AI Agent Sessions - Restores previously closed AI agent sessions and returns agents to their last active state.
• Active Session Visibility - Provides a sidebar that displays all current terminal sessions for fast navigation and tracking.
• Agent Attention Alerts - Uses visual alerting rings and unread indicators to signal when AI agents require user decisions.
• Sidebar Log Displays - Displays terminal commands and tool logs in a right-sidebar for non-intrusive monitoring.
• Terminal Session Mirroring - Streams live terminal output and notifications from a desktop to a mobile device for remote monitoring.
• DOM Manipulation - Integrates a web engine for direct programmatic manipulation of the DOM and session state within embedded surfaces.
• Browser Control APIs - Exposes APIs for programmatic control over the integrated browser, including accessibility tree capture and JS evaluation.
• Agent Lifecycle Management - Arranges agent panes in a grid and removes idle agents automatically after a period of inactivity.
• Automated Surface Management - Automatically closes idle agent panes and queues new requests when screen space is unavailable.
• Agentic Workflow Orchestrators - Executes specialized agent roles and parallel team tasks via scriptable terminal workflows.
• Session Archiving - Indexes coding agent transcripts with full-text search for reviewing past conversations.
• Multi-Model Orchestration - Spawns parallel agents using different language models, mapping each to a unique terminal split.
• Agent Resource Monitoring - Provides tracking of CPU and RAM usage specifically for active AI coding agents via a task manager.
• Session Search - Provides tools for searching and retrieving archived AI agent sessions from a vault into the active workspace.
• Workspace Group Management - Offers programmatic control over the organization, visibility, and styling of terminal workspace groups via CLI.
• DOM State Inspection - Retrieves DOM attributes, visibility states, and captures screenshots for programmatic page analysis.
• Browser Session Data Management - Provides the ability to read, write, and restore cookies and local storage to maintain authenticated browser states.
• Application State Restorations - Recreates window layouts, working directories, and browser history after an application restart.
• Custom Command Definitions - Allows for the definition of reusable shell commands and agent triggers accessible via the command palette.
• Custom Command Execution - Executes custom bash scripts to automate recurring tasks such as managing git worktrees or SSH sessions.
• Notification-Triggered Execution - Automatically executes specified shell commands in response to the scheduling of system notifications.
• AI Session Synchronization - Synchronizes the terminal environment with external AI coding agents using specialized hooks and binary paths.
• Project Metadata Displays - Displays real-time project information, including git branches and listening ports, within the sidebar.
• Prompt Input Interfaces - Provides a dedicated editor surface for drafting and refining AI prompts before submission.
• Shell Command Execution - Triggers and controls the execution of operating system shell commands within focused terminal panes.
• Surface Focus Histories - Provides a searchable, timestamped list of all previously closed and focused surfaces for recovery.
• Multiplexer Mapping - Translates external terminal multiplexer commands into native workspace API calls for agent rendering.
• Terminal Automation - Automates terminal workflows including workspace creation, input delivery, and screen reading via CLI and socket.
• Input Injection - Provides programmatic delivery of text strings and key presses to terminal surfaces via a control plane.
• Layout Automation - Automates the creation of specific workspace configurations with split-pane terminals and browser windows via CLI.
• Surface Tab Stacking - Groups different types of terminal and browser tabs within a single split pane.
• Workspace Command Orchestration - Executes commands scoped to the active workspace and surface to add panes or inspect state.
• Path-Based Grouping - Automatically organizes the sidebar and workspaces based on filesystem path prefixes.
• Visual Workspace Grouping - Provides the ability to cluster related workspaces into collapsible, named sections with custom icons and colors for better organization.
• Remote Workspace Control - Allows switching between terminal workspaces and navigation tabs from a mobile interface.
• Workspace Layout Restorations - Reopens recently closed terminal tabs, browser tabs, and workspaces while preserving internal pane layouts.
• Mobile Push Notification Deliveries - Forwards alerts from terminal panes to mobile devices using push notification services.
• Device Notification Bridges - Routes alerts from AI agents running on remote machines directly to the local sidebar.
• Agent Telemetry Streams - Synchronizes AI agent activity and execution events with external tools through dedicated telemetry streams.
• SSH Client Connections - Implements SSH client connectivity to remote hosts for executing commands and integrating notifications.
• Traffic Proxying - Routes HTTP and WebSocket traffic through a remote machine to allow access to development servers.
• SSH - Transmits terminal data and commands over SSH using a background daemon to maintain connectivity.
• Session State Reconstitution - Uses resume tokens and metadata to reconstitute window layouts and active agent processes after restart.
• Task-Based - Creates, lists, selects, and closes isolated workspaces to organize concurrent tasks.
• DOM State Synchronization - Blocks automation execution until specific DOM selectors, text fragments, or URL patterns are detected.
• Notification Interception Scripts - Provides programmable hooks and JSON filters to modify or filter notifications before they are displayed.
• Activity Execution Logs - Maintains a timestamped sidebar feed of log entries with severity levels to track agent and build activity.
• Event-Driven Notification Triggers - Triggers visual and auditory cues across workspaces based on intercepted system alerts via shell hooks.
• Remote-to-Local Alerting - Displays alerts in the local sidebar when processes on a remote host trigger notification commands.
• Element Interaction - Implements high-level APIs to simulate user interactions like clicking, typing, and scrolling on web elements.
• Process Completion Alerts - Triggers visual badges and audible notifications immediately upon the completion of background processes.
• Input Bars - Provides a specialized text entry interface for sending commands and shortcuts to remote sessions.
• Notification History Centers - Provides a dedicated history panel for viewing past alerts and jumping directly to the associated workspace.
• Progress Indicators - Displays numerical progress bars with custom labels in the sidebar for tracking long-running jobs.
• Status Indicators - Utilizes custom status pills, labels, and icons in the sidebar to track the state of integrated tools.
• Terminal Escape Notifications - Triggers native desktop alerts via CLI, escape sequences, or scripts to notify users of background tasks.
• JavaScript Injections - Provides capabilities for injecting custom JavaScript and styles directly into the embedded browser environment.
• Focus History Navigation - Implements linear navigation controls to move through a chronological stack of previously focused windows and workspaces.
• Unread State Navigation - Tracks and navigates to active workspaces using unread markers to prioritize agent attention.
• Browser Dialog Handlers - Programmatically accepts or dismisses native browser dialogs and alerts during automation.
• Browser Navigation - Directs the embedded browser to specific URLs and manages page load events and navigation history.
• Remote Browser Connections - Proxies browser panes to ensure local requests reach a remote server network.

Star history