VS
SimDrive vs Maestro
Maestro is great for human-authored, YAML-driven flows. SimDrive optimizes for an LLM agent calling Python functions. Same simulator underneath; different ergonomics on top.
| SimDrive | Maestro | |
|---|---|---|
| Primary caller | LLM agent | Human via YAML |
| Input | Python / MCP | YAML flow file |
| Observation | screenshot + a11y tree | hierarchy dump |
| Record & replay | first-class with SSIM drift gates | Studio recorder, selector-match replay |
| MCP server | vision-first primitives (observe + act) | built-in since Feb 2026, YAML-flow oriented |
| Drift detection | SSIM + state contract | selector match |
| License | Elastic-2.0 | Apache-2.0 |
When to use Maestro
You have a stable iOS app, a QA team that writes test flows by hand, and you want a YAML DSL with predictable selectors. Maestro is a mature, well-engineered choice for that use case.
When to use SimDrive
You want an AI agent — Claude, GPT, a local model — to drive the simulator directly. You want structured state going into the model and structured actions coming out, without inventing a selector DSL. You want deterministic replays you can run free in CI, with SSIM-based drift detection that fails fast when the UI changes underneath a recording.
Maestro shipped a built-in MCP server in February 2026; it focuses on generating and running
YAML flows. SimDrive's MCP is shaped around two vision-first primitives (observe +
act) with replay as a first-class artifact, designed for agents that author the
test live rather than running a pre-written flow.