Drive the
iOS Simulator
with Claude.

PRIMITIVES

Three functions. That's the whole API.

No DSL. No YAML. No selectors. Your agent reads the simulator state and acts on it the same way a human tester does — only deterministically.

01

ios_observe()

Screenshot + accessibility tree, returned as a typed state object.

from simdrive import ios_observe

# Capture the current simulator state — screenshot + accessibility tree.
state = ios_observe()
print(state.elements)  # list of (role, label, frame) tuples

02

ios_act()

Claude observes the screen and acts via a small, sharp tool surface — tap, type, swipe. Resolve targets against the last observation when you name them.

from simdrive import ios_act

# Drive the simulator. When you name a target (mark id, label, or
# stable_id), the call is resolved against the last observation.
# Raw (x, y) coordinates are accepted too as an escape hatch.
ios_act("tap", target="Sign In")
ios_act("type", text="hello@example.com")

03

ios_record()

Capture a deterministic replay. Run it free in CI — no AI tokens.

from simdrive import ios_record

# Capture a deterministic replay file as the agent works.
with ios_record("smoke.replay.json"):
    ios_act("tap", target="Sign In")
    ios_act("type", text="hello@example.com")
# Replay it free, forever, in CI.

HOW IT WORKS

Observe → Act → Record. Then replay.

COMPARISON

Why not Maestro or Appium?

Both are excellent for human-authored tests. SimDrive optimizes for a different caller: an LLM agent that needs structured state in and structured actions out.

MCP INSTALL

Plug into your agent in one config block.

SimDrive ships an MCP server. Add it to your client config — nothing else to install.

# .mcp.json — Claude Code
{
  "mcpServers": {
    "simdrive": {
      "command": "simdrive"
    }
  }
}

# ~/.cursor/mcp.json — Cursor
{
  "mcpServers": {
    "simdrive": {
      "command": "simdrive"
    }
  }
}

# ~/.continue/config.json — Continue
{
  "experimental": {
    "modelContextProtocolServers": [
      { "transport": { "type": "stdio", "command": "simdrive" } }
    ]
  }
}

PRICING

Free to run. Pay to scale.

Replays are always free. Pricing covers parallel CI, team seats, and priority support.