Maestri is a native macOS app that gives you an infinite spatial canvas to orchestrate multiple AI coding agents like Claude Code, Codex, and OpenCode. Each terminal becomes a visual node you can position, connect, and coordinate through PTY-based agent-to-agent communication. Add markdown notes, sketches, and files to build persistent shared memory across sessions. Built with Swift and Metal for smooth performance, it targets developers running parallel agents who want a single visual workspace to plan, monitor, and preserve project context.

• Infinite Spatial Canvas: GPU-accelerated Metal canvas where terminals, notes, and sketches live as draggable nodes. Pan and zoom without frame drops while keeping RAM available for your agents.
• Agent-to-Agent PTY Orchestration: Connect terminals with a line and agents talk to each other directly through PTY. Works across any CLI agent, no APIs or middleware needed.
• Ombro On-Device Companion: An Apple Foundation Models-powered assistant that watches your agents, summarizes completed tasks, and flags conflicts like two agents editing the same file. Runs fully local with no cloud calls.
• Shared Markdown Notes: Connect agents to sticky notes saved as real markdown files on disk. Multiple agents can read and write to the same note, creating shared memory that persists across sessions and harnesses.
• Custom Agent Roles: Define personas like Conductor, Reviewer, or Web Researcher with color badges and icons. Assign them at terminal launch instead of rewriting CLAUDE.md every time.
• Multi-Workspace Management: Create independent workspaces per project with icons, colors, and target directories. Switch between them with tmux-style shortcuts and organize into pinned groups.
• APFS Workspace Cloning: Instantly clone a workspace at near-zero cost using copy-on-write. Test fixes, review PRs, or try experimental branches without touching your main working tree.
• IDE Integration: Direct hooks into VS Code, Zed, and Xcode so you can jump from the canvas into deeper code work.

• Privacy-First Design: Local-only architecture with no telemetry and on-device AI through Apple Foundation Models. Your project data never leaves your Mac.
• Native Performance: Built entirely in Swift and SwiftUI with Metal rendering. No Electron overhead, smooth panning and zooming even at heavy zoom levels.
• Real Multi-Agent Workflow: PTY-level communication actually makes different CLI agents cooperate, not just run in parallel tabs.
• Persistent Context: Markdown notes on disk act as durable shared memory across agents and sessions, solving a real pain point in multi-agent coding.
• Affordable Pro Plan: Pro is affordable for solo developers compared to typical subscription tooling, with support for up to two Macs.

• macOS Only: Requires Apple Silicon and a recent macOS version. Windows and Linux users are excluded entirely.
• Not a Full IDE: You will still switch to VS Code, Zed, or Xcode for deep code editing and debugging.
• Agents Sold Separately: You still pay for Claude Code, Codex, or API usage on top of Maestri. Real monthly cost depends on which agents you run.
• Learning Curve: Spatial canvas plus role definitions plus PTY connections takes time to set up before the productivity gains show up.

• Parallel Feature Development: Run one agent building a feature while another writes tests and a third handles documentation. Connect them so context flows between terminals without manual copy-paste.
• PR Review with Isolated Branches: Clone your workspace with APFS copy-on-write to review a pull request in isolation. Keep your main work untouched while an agent walks through the diff.
• Multi-Agent Research and Coding: Assign one agent a Web Researcher role and another a Coder role. The researcher pulls context into a shared markdown note, the coder reads it and implements.
• Long-Running Project Context: Use shared markdown notes as persistent memory across weeks of work. Agents document decisions, and future sessions pick up where past ones left off.
• Solo Developer Orchestration: Manage several projects on one canvas with separate workspaces. Ombro summarizes what happened while you were away so you can jump back in quickly.
• Experimental Refactoring: Clone a workspace instantly to test a risky refactor with an agent. Merge the result cleanly if it works, discard if it does not.