Orchard supervises AI agents across code, simulators, browsers, backends, and release pipelines. Every task produces evidence. Every release has proof.
9 gates. Machine-checkable contracts.
Evidence for every action.
Other tools help agents write code. Orchard proves the code actually works.
TriggerEvery code change
TriggerEvery build success
TriggerUI changes
TriggerUI changes
TriggerAuth/API changes
TriggerLaunch/render changes
TriggerAny UI change
TriggerEvery commit
TriggerContract-defined
Every completed task produces a Case — a git-committed investigation record with structured evidence, agent transcripts, contract results, and cost attribution.
Proofs, not vibes.
| Method | Path | Status | Time |
|---|---|---|---|
| POST | /auth/apple | 200 | 142ms |
| GET | /user/profile | 200 | 89ms |
| GET | /user/settings | 200 | 67ms |
| POST | /analytics/event | 204 | 34ms |
Built for Agents, Not Extensions
Every component is purpose-built for autonomous Apple development
Deterministic state machine governs every agent session
8 specialized roles, one coordinated pipeline
| Server | Tools |
|---|---|
| XcodeBuildMCP | 79+ |
| Apple mcpbridge | ~20 |
| mobile-mcp | 40+ |
| Playwright | ~30 |
Deterministic state machine governs every agent session
8 specialized roles, one coordinated pipeline
| Server | Tools |
|---|---|
| XcodeBuildMCP | 79+ |
| Apple mcpbridge | ~20 |
| mobile-mcp | 40+ |
| Playwright | ~30 |
Four layers. One binary.
Zero configuration.
Canvas OS
GPUI Shell + Workspace Runtime
A GPU-accelerated native macOS shell built on GPUI and Metal, rendering the entire workspace at 120 fps. Every panel, canvas, and overlay is a retained-mode GPU surface — no Electron, no DOM, no compromise.
Not just a code editor. A development operating system.
Three platform capabilities that no other tool offers.
Workflow Automation
Visual workflow engine with 400+ integrations. Automate TestFlight distribution, crash triage, screenshot baselines, and 10 more pre-built templates.
Provider Picker
Choose your LLM per agent role. Toggle MCP servers. Set verification gates. Three tiers of progressive disclosure — from one-click to raw YAML.
Cost Management
Per-slice budgets with progressive throttling. Smart routing picks the cheapest model that meets capability requirements. Every case includes cost attribution.
Every user interaction feeds the growth engine. Zero manual intervention.
PostHog events → n8n workflows → Customer.io lifecycle → Resend delivery
Built on proven foundations.
Built on proven foundations.
Composed, not invented.
GPUI
Apache 2.0GPU-accelerated UI framework from Zed
Rust
MIT / ApacheSystems language for the kernel binary
Tokio
MITAsync runtime for concurrent task execution
Claude Code
ProprietaryAnthropic's agentic coding assistant
Codex CLI
Apache 2.0OpenAI's terminal-native coding agent
Gemini CLI
Apache 2.0Google's multimodal coding agent
XcodeBuildMCP
MIT79+ Xcode build & project management tools
mobile-mcp
MIT40+ iOS simulator & device interactions
mcpbridge
MIT~20 Apple system bridge tools
Kernel
ProprietaryCore orchestration and process management
SQLite
Public DomainEmbedded database for state and evidence
FTS5
Public DomainFull-text search across evidence and logs
Structured for Agent Workflows
Feature Slices
Evidence
Releases
“The world does not need another AI code editor. It needs a development operating system that can supervise agents across code, simulators, browsers, backends, and release pipelines — with proofs, not vibes.”
Watch agents work in real time.
Orchard Alpha 0.1
Private preview starting Q3 2026. Join the waitlist.
Evidence Verification
Scanning artifacts...
Activity Feed
LIVEFeature: Add dark mode to Settings
Implementer → Claude Code 4.5
Debug build in 3.2s
14/14 (2 new)
4 artifacts, 3 screenshots
All 9 gates passed ✓
Slice #7 Complete
All evidence verified. 14/14 tests, 0 a11y violations, 42MB memory peak.
Agent Capacity
3 of 4 agent slots active. Queue depth: 2 slices.
Terminal-native. Agent-compatible. CI-ready.
Full control from the command line. Agents use the same vocabulary.
CI/CD Integration
Exit 0 = all gates pass
Agent Interop
ORCHARD_API env var auto-set in managed terminals. Claude Code, Codex, Gemini can read slice state and attach evidence.
SDK
orchard-sdk (Rust crate) + OrchardKit (Swift package). Custom gates, providers, evidence collectors. Hot-reload.
72 tools. 17 categories. All managed.
MCP Servers
Connected tool providers
| Server | Tools | Protocol | Status |
|---|---|---|---|
| XcodeBuildMCP | 79+ | stdio | Active |
| Apple mcpbridge | ~20 | stdio | Active |
| mobile-mcp | 40+ | stdio | Active |
| Playwright MCP | ~30 | stdio | Active |
| DesignReview MCP | 5 | stdio | Active |
Agent Roles
Orchestrated task execution
| Role | Provider | Capabilities | Avg Time |
|---|---|---|---|
| Planner | Claude Code | Architecture, task decomposition | 45s |
| Implementer | Claude Code | Code generation, refactoring | 2m |
| Verifier | Codex CLI | Testing, assertion | 30s |
| Mobile Runner | mobile-mcp | Simulator, device testing | 1m |
| A11y Auditor | Claude Code | VoiceOver, accessibility | 20s |
| Health Reviewer | Claude Code | Performance, memory | 15s |
| Evidence Collector | Orchard Core | Screenshots, reports | 10s |
| Release Manager | Orchard Core | Contract gates, deploy | 5s |
Tool Categories
Coverage by domain
| Category | Count | Coverage | Trend |
|---|---|---|---|
| Build & Compile | 12 | 100% | ↑ |
| Testing | 18 | 94% | ↑ |
| Signing & Deploy | 8 | 100% | → |
| UI Automation | 15 | 87% | ↑ |
| Accessibility | 5 | 100% | ↑ |
| Performance | 6 | 92% | ↑ |
| Evidence & Audit | 8 | 100% | → |
How Orchard is different.
The only tool that proves your iOS app works end-to-end.
| Capability | Cursor | Trellis | Augment | Warp | Orchard |
|---|---|---|---|---|---|
| Apple-native execution | ✘ | ✘ | ✘ | ✘ | ✔72+ tools |
| Verification gates | ✘ | ✘ | ✘ | ✘ | ✔9 gates |
| Evidence pipeline | ✘ | Partial | ✘ | ✘ | ✔Full |
| Contract engine | ✘ | ✘ | ✘ | ✘ | ✔Machine-checkable |
| Multi-agent roles | ✘ | 1 (Claude) | ✘ | ✘ | ✔8+1 roles |
| Runtime slices | ✘ | Worktrees | ✘ | ✘ | ✔Full isolation |
| Workflow automation | ✘ | ✘ | ✘ | ✘ | ✔n8n + 10 templates |
| Release pipeline | ✘ | ✘ | ✘ | ✘ | ✔TestFlight + ASC |
| Cost management | ✘ | ✘ | ✘ | ✘ | ✔Per-slice budgets |
| CLI tool | ✘ | ✔trellis-ctl | ✘ | ✔ | ✔orchard CLI |
Orchard isn't competing with code editors. It's building the verification and release layer that sits above them.
Zero-config by default. Endlessly configurable.
Get Started
brew install orchard-ide/tap/orchardJoin the founding team.
Orchard is pre-seed. We're looking for a technical co-founder, seed investors, and early design partners.
Co-Founder
Looking for a Rust/GPUI or iOS/Swift expert to build the core IDE. Equity-heavy, salary-light. Must ship fast and think in systems.
Seed Investment
Pre-seed round to fund 2 engineers for 12 months. Building the agent-native IDE layer Apple developers deserve.
Design Partner
iOS teams shipping weekly who want to 10x velocity with supervised agents. Early access + direct founder support.
TAM
$52B
AI agent market by 2030
Beachhead
2.4M
Apple developers worldwide
Pricing
$20
/seat/month (Pro tier)
Timeline
16mo
to v1.0 public launch
Early Access
We'll notify you when early access begins.
Built from the trenches.
Rory is a full-stack developer and serial founder who has spent years building production software with AI coding agents — and experiencing firsthand every pain point Orchard solves.
As founder of Auron Intelligence (clinical AI for vascular imaging), ePrescience (AI-powered learning platform), and wemakepages.com (web design agency), he manages AWS infrastructure, multi-platform builds, and complex agent workflows daily.
Orchard was born from the realization that the iOS build-test-iterate-ship loop is uniquely painful — and uniquely solvable — for agentic development. After hundreds of hours directing Claude Code, Cursor, and Codex on real iOS projects, the pattern became clear: the tooling above the editor is missing.