AgentSwift

Download AgentSwift-0.1.zip You must install the dependencies listed below in order for the binary to work. See below for setup commands.

Dependencies:

Xcode

Xcode command line tools

xcodebuildmcp

openspec

A native macOS app that runs an autonomous AI coding agent for Apple platform development. Describe what you want to build, and AgentSwift uses Claude to discover your project, implement changes, build, run, and validate — without you touching Xcode.

What it does

AgentSwift drives a multi-phase agentic workflow:

Discover — Claude inspects your Xcode project structure and schemes Implement — edits source files to match your request Build — runs xcodebuildmcp to compile Launch / Validate — boots the app on a simulator or macOS, runs UI automation to verify behavior Archive — marks the task complete

Requirements

macOS 26.1+

Xcode

Node.js / npm

An Anthropic API key

Dependencies

Install these two CLIs before running the agent:

xcodebuildmcp

Provides build, launch, and UI automation capabilities for Xcode projects.

npm install -g xcodebuildmcp

openspec

Tracks implementation specs across agent sessions.

npm install -g @fission-ai/openspec

Setup

Build and run the app in Xcode. Open Settings and enter your Anthropic API key. Select a Project Folder (the root of your Xcode project). Optionally pick an iOS Simulator from the dropdown. Type what you want to build and press Cmd+Return.

On the first run the agent discovers your project's scheme and simulator target. Subsequent runs skip discovery and go straight to implementation.

Models

Model Use when Claude Opus 4.7 Complex tasks, large codebases Claude Sonnet 4.6 Faster iteration, lighter tasks

Key behaviors

Message queuing — if you send a new message while the agent is running, the latest supersedes earlier ones

— if you send a new message while the agent is running, the latest supersedes earlier ones Build caching — scheme, project path, and simulator ID are extracted after the first build and reused automatically

— scheme, project path, and simulator ID are extracted after the first build and reused automatically Error escalation — the agent attempts one fix on a failure, then surfaces the error to you rather than looping

Architecture

AgentSwiftApp.swift — app entry point ContentView.swift — UI, view models, agentic loop AnthropicService.swift — Anthropic API client (streaming SSE) ToolExecutor.swift — bash / read_file / write_file execution Item.swift — chat message model

No external Swift dependencies — pure SwiftUI + Foundation.