Scape

Product Introduction

1. Definition: Scape is a local-first, macOS-native command center and integrated development environment (IDE) designed specifically for orchestrating multiple AI coding agents. It functions as a terminal replacement that integrates isolated Git worktrees, a code editor, a local database, and automation tools into a single visual workspace.
2. Core Value Proposition: It exists to solve the cognitive overload and technical friction of running multiple AI agent sessions in parallel. Scape enables developers, engineers, and product builders to scale their output by commanding a "fleet" of AI agents (like Claude Code, Codex, or OpenCode) on their local machine, building owned automations, and maintaining full control over their infrastructure and intellectual property.

Main Features

1. Parallel, Isolated Git Worktrees: This is the core technical architecture. Each AI agent session runs in its own isolated Git worktree—a separate branch with an independent working directory. This allows multiple agents to work on the same codebase simultaneously without file conflicts. The system manages the Git operations, letting users spawn new worktree sessions with one click. It intelligently handles .gitignore rules, requiring toolkit scripts for environment-specific files.
2. Playbooks & Tables (Local Automation Infrastructure): Playbooks are executable runbooks that chain bash commands, HTTP requests, and AI-powered steps to automate complex workflows (e.g., releases, data collection, ticket triage). Tables are a built-in, local database (syncing via the user's iCloud) that Playbooks can populate. This creates a compounding, owned automation system that runs entirely on the user's Mac, with no external server dependency.
3. Argus Orchestration & Watchdogs: Argus is a meta-agent designed to manage other agents. Users give Argus a high-level objective and guardrails; it then spins up and manages child agent sessions, handles routine approvals, and escalates only critical decisions. Watchdogs are specialized sessions that monitor other active agent sessions, automatically responding to prompts or questions based on predefined instructions, acting as an intermediate layer between full manual control and blind acceptance.
4. Multi-Agent Collaboration & Encrypted Comms: Supports two collaboration modes: Agent ↔ Agent Chat, where two different AI models (e.g., Claude vs. Codex) debate or collaborate within a single local session, and Agent Rendezvous, which connects a user's local agent to another user's agent over the network for cross-machine planning. It also provides end-to-end encrypted chat rooms for human collaboration, with messages relayed as ciphertext and no persistent history.

Problems Solved

1. Pain Point: The inefficiency and context-switching cost of managing multiple AI coding agent sessions in separate terminal windows or IDE tabs. Traditional workflows cap output at one agent, one prompt, one wait cycle.
2. Target Audience: Software engineers, engineering managers, indie developers, product builders, and technical creators who extensively use AI coding assistants (Claude Code, Cursor, GitHub Copilot) and need to scale their ideation, review, and automation processes. It's for users who have "a lot of ideas and not enough time."
3. Use Cases:
   • Parallel Feature Development: Simultaneously refactoring a payments system with one agent while another drafts documentation and a third fixes bugs.
   • Automated Workflow Management: Building a Playbook that fetches overnight emails, uses an AI agent to flag VIP messages and draft replies, and files the results into a local Table for review.
   • Orchestrated Code Reviews: Using Argus to manage a fleet of agents performing automated testing, linting, and security analysis on a pull request, summarizing only the critical issues.
   • Cross-Team Agent Collaboration: A developer's agent directly interfacing with a designer's agent via Rendezvous to plan API contracts and UI component specifications.

Unique Advantages

1. Differentiation: Unlike cloud-based AI platforms or simple IDE plugins, Scape is a local-first, comprehensive workspace that doesn't proxy or store user code/data. It contrasts with traditional terminals by adding visual session management, and with full IDEs by focusing on agent orchestration over manual coding. Competitors typically offer single-session interfaces or cloud-dependent automation.
2. Key Innovation: The integration of isolated Git worktrees as a first-class primitive for AI agent sessions. This technical approach provides a clean, conflict-free, and git-native method for true parallelism. Combined with the local automation stack (Playbooks/Tables) and the orchestration layer (Argus/Watchdogs), it creates a closed-loop system where agents can build and use tools that permanently enhance the user's local workflow, compounding in value.

Frequently Asked Questions (FAQ)

1. Does Scape send my code or API keys to its servers? No. Scape is a local-first application for macOS. All code, AI model interactions (using your own subscriptions), API keys, session data, and automation logic execute and reside solely on your machine. Notes sync via your personal iCloud account.
2. How does Scape's pricing model work, and what happens if I cancel? Scape offers a one-time purchase for a month of updates and a Pro/Supporter subscription. If your subscription ends, all versions released during your active period continue working permanently. Only subscription-locked features (Playbooks, Tables, Agent Rendezvous, Encrypted Chat) become read-only; core features like worktrees, watchdogs, and the editor remain fully functional.
3. Can I use Scape with any AI coding assistant? Scape currently provides first-class "harnesses" for Claude Code, OpenAI Codex (via Cursor or Copilot), and OpenCode. You must bring your own active subscription to these services. The architecture allows per-session harness selection, enabling comparison and debate between different AI models within the same workspace.
4. What are the system requirements for running multiple AI agents locally? Scape runs on macOS and leverages the machine's local resources. Running many parallel agents with large context windows will significantly increase RAM and CPU usage, as each agent session runs its own process. The integrated local voice transcription uses Apple's Neural Engine via CoreML for efficient offline processing.
5. Is Scape suitable for team or enterprise use? While currently focused on individual productivity, features like Agent Rendezvous and Encrypted Chat Rooms enable secure cross-machine collaboration. The local-only data stance may present deployment challenges for enterprises requiring centralized management, but it offers superior privacy and control for individual professionals and small teams.