SpecterQA iOS
The only iOS tester your agent can call.
Overview
SpecterQA iOS is an AI-native iOS simulator testing tool built around one insight: use AI to record, use a deterministic engine to replay.
You pay for AI tokens exactly once. Claude drives your iOS simulator using vision and Set-of-Mark prompting — tapping elements accurately, filling forms, navigating flows — and saves every action as a deterministic YAML replay file. Every subsequent CI run executes that replay file without ever calling Claude. No AI tokens. No API cost. The same result, every time.
The MCP server exposes 19 tools, making SpecterQA natively callable from Claude Code, Claude Desktop, or any MCP-compatible orchestrator. It is also A2A discoverable via `.well-known/agent.json`. Your AI agent can boot a simulator, record a smoke test, and queue it for CI — all in a single session.
Maestro users get a zero-cost migration path: existing YAML files work natively. `tapOn`, `inputText`, `assertVisible`, `assertNotVisible`, and `waitFor` are all understood without changes. Mix Maestro shorthand and SpecterQA native syntax in the same file.
BYOK means your Anthropic API key stays with you. SyncTek never proxies it, stores it, or sees your test data. Your app binary, simulator state, and recordings stay on your machine.
Features
Record Once, Replay Free
AI records test sessions using Claude vision. Deterministic engine replays in CI — zero API cost on reruns.
Agent-Native (19 MCP Tools)
Full testing surface exposed as MCP tools. Claude Code, Claude Desktop, or any MCP-compatible orchestrator can drive iOS tests in plain English.
Maestro Compatible
Existing Maestro YAML files work natively. tapOn, inputText, assertVisible — all understood. Zero migration cost.
Parallel CI — 10x Faster
Shared runner reuse and clone isolation. Run 10 replays in the time XCUITest runs 1. `--parallel N` flag.
90% Tap Accuracy
Set-of-Mark (SoM) prompting annotates the simulator screenshot before Claude taps. No coordinate guessing, no brittle selectors.
BYOK — Full Data Control
Bring your own Anthropic API key. SyncTek never sees it. Your test data never leaves your machine. 97% gross margin for us, full control for you.
Quick start
Install CLI + MCP server
$ pip install 'specterqa-ios[mcp]' Verify Xcode + simulators
$ specterqa-ios setup Scaffold .specterqa/ project files
$ specterqa-ios init Set your API key (record phase only)
$ export ANTHROPIC_API_KEY=sk-ant-... Record a test with Claude
$ specterqa-ios run --product myapp --journey smoke Replay all tests in CI (free)
$ specterqa-ios ci .specterqa/replays/ Pricing
Trial
Free
1 simulator, 3 runs/session. No credit card required. Get started immediately.
Indie
$29/mo
2 simulators, unlimited runs. Solo developers and side projects.
Pro
$99/mo
4 simulators, unlimited runs, parallel CI. The workhorse tier.
Team
$299/mo
10 simulators, parallel CI, priority support. Built for mobile teams.
Enterprise
Custom
Unlimited simulators, SLA, dedicated support, custom contracts.
Cost transparency
Every run costs money (Anthropic API calls). We built cost control directly into the engine so there are no surprises.
| Journey type | Typical cost |
|---|---|
| Initial record (smoke, ~5 taps) | $0.05 - $0.15 |
| Initial record (full journey, ~20 taps) | $0.20 - $0.60 |
| CI replay (any size) | $0.00 |
Honest tradeoffs
- macOS + Xcode 15+ required — iOS Simulator is macOS-only. No Linux CI without a macOS runner.
- BYOK required — you must supply your own Anthropic API key for the record phase. Replay is free.
- Elastic License 2.0 — free for internal use; commercial redistribution or competing service use is restricted.
- Record phase requires network access to the Anthropic API. Replay is fully offline.
- Parallel CI (Pro tier and above) — Trial and Indie tiers run replays sequentially.
- Simulator-only — physical device testing is not currently supported.