Catenary

Product Introduction

1. Definition: Catenary is a local-first, spatial Integrated Development Environment (IDE) and terminal orchestrator designed specifically for managing and collaborating with multiple AI coding agents. It is a desktop application built on Electron, providing an infinite canvas workspace.
2. Core Value Proposition: It exists to solve the terminal and context management chaos inherent in modern AI-assisted development. Catenary enables developers to visually orchestrate AI agents like Claude Code, Cursor, and Codex within a unified, private workspace, eliminating tab overload and manual context passing.

Main Features

1. The Infinite Spatial Canvas: The core interface is a zoomable, pannable infinite canvas where native PTY terminals, Monaco code editors (with integrated Git diffing), browser previews, and agent panels coexist as movable panels. This spatial arrangement allows for simultaneous monitoring of all active AI agents and processes.
2. Visual Context Wires: This feature enables directed data flow between panels. Users can drag a visual cable from a terminal output or editor to an agent's input. Context can be relayed automatically or on-demand, creating a clear, visual pipeline for passing prompts, errors, and code snippets between different AI models and tools without copy-pasting.
3. 1-Click Isolated Task Islands: With a single click, users can spawn a new "Task Island"—an isolated Git worktree on a new branch with its own working directory and a dedicated AI agent. This allows for parallel experimentation (e.g., testing three different refactors) without contaminating the main working tree. Each island is color-coded and maintains its state.
4. Maestro Mode (Agent Orchestration): An overseer AI agent can be activated to recruit, brief, and wire a team of subordinate helper agents (2-10, configurable). Every action proposed by the Maestro agent requires explicit human approval via an editable card before execution, maintaining developer oversight over automated multi-agent workflows.
5. Local-First & Zero-Telemetry Architecture: The application is designed for complete offline operation. All code, terminal data, and prompts remain on the user's machine. Network traffic is strictly limited to update checks (GitHub) and license activation (thecatenary.app), with all analytics and crash reporting endpoints physically removed from the distributed build.

Problems Solved

1. Pain Point: Context fragmentation and terminal sprawl when using multiple AI coding assistants. Developers struggle to track concurrent agent sessions, manually copy context between windows, and manage parallel coding experiments.
2. Target Audience: Software engineers and developers who regularly use AI coding assistants (Claude Code, Cursor, GitHub Copilot) for complex tasks like refactoring, debugging, and feature development. It is particularly valuable for full-stack developers, tech leads, and indie hackers managing multi-faceted projects.
3. Use Cases: Orchestrating a multi-agent squad where one agent writes code, another reviews it, and a third runs tests. Safely exploring multiple solution branches for a bug fix simultaneously. Maintaining visual oversight of a long-running AI-driven refactoring session across an entire codebase while continuing other work.

Unique Advantages

1. Differentiation: Unlike traditional IDEs or cloud-based AI platforms, Catenary provides a spatial, multi-panel workspace specifically for agent orchestration. It differs from mere terminal multiplexers by integrating native editors, visual wiring, and Git-aware task isolation directly into the canvas.
2. Key Innovation: The combination of a local-first, zero-telemetry guarantee with visual agent-context wiring and one-click Git worktree isolation. This trio ensures privacy, clarifies the flow of information between AI tools, and safely enables high-velocity parallel development, all within a single visual context.

Frequently Asked Questions (FAQ)

1. Is Catenary a cloud service or a local application? Catenary is a 100% local-first desktop application for macOS, Windows, and Linux. Your code, API keys, and AI prompts never leave your machine, as it has no backend server to send them to.
2. How does Catenary handle my API keys for services like OpenAI or Anthropic? API keys are stored locally on your device and are used directly by the agent CLI (e.g., cursor, claude) you configure. Catenary itself does not proxy or have access to your keys; communication is directly between your machine and the AI provider's API.
3. Can I use Catenary with locally-run AI models like Ollama? Yes, Catenary is compatible with locally-hosted models. You can configure your agent CLI to point to a local endpoint (e.g., localhost:11434 for Ollama), and all interactions will remain entirely on your network.
4. What are the limitations of the free version versus the Pro license? The free version limits users to one open project at a time and two parallel tasks (Task Islands). The Pro license removes these limits, allows unlimited parallel tasks and projects, unlocks the full Maestro Mode team size, and permits activation on up to 2 personal machines.
5. How does the visual context wiring work technically? Wires create a directed data channel between panels. When activated, they capture the standard output (stdout) or selected content from a source panel (like a terminal) and pipe it as input to a target panel (like an AI agent's prompt interface). This is managed internally via IPC (Inter-Process Communication) within the Electron app.